Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance

Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance Sam Jones, Eva Maria Hodel, Raman Sharma, Katherine Kay, Ian M Hastings Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 The Journal of Infectious Diseases MAJOR ARTICLE Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance 1, 1,2,a 1 1,b 1 Sam Jones, Eva Maria Hodel, Raman Sharma, Katherine Kay, and Ian M. Hastings 1 2 Parasitology Department and Department of Clinical Sciences, Liverpool School of Tropical Medicine, United Kingdom (See the Editorial Commentary by Small and Seydel on pages 1176–7.) Background. Standard treatment for severe malaria is with artesunate; patient survival in the 24 hours immediately posttreat- ment is the key objective. Clinical trials use clearance rates of circulating parasites as their clinical outcome, but the pathology of XX severe malaria is attributed primarily to noncirculating, sequestered, parasites, so there is a disconnect between existing clinical metrics and objectives. XXXX Methods. We extend existing pharmacokinetic/pharmacodynamic modeling methods to simulate the treatment of 10 000 patients with severe malaria and track the pathology caused by sequestered parasites. Results. Our model recovered the clinical outcomes of existing studies (based on circulating parasites) and showed a “sim- plified” artesunate regimen was noninferior to the existing World Health Organization regimen across the patient population but resulted in worse outcomes in a subgroup of patients with infections clustered in early stages of the parasite life cycle. This same group of patients were extremely vulnerable to resistance emerging in parasite early ring stages. Conclusions. We quantify patient outcomes in a manner appropriate for severe malaria with a flexible framework that allows future researchers to implement their beliefs about underlying pathology. We highlight with some urgency the threat posed to treat- ment of severe malaria by artemisinin resistance in parasite early ring stages. Keywords. Plasmodium falciparum; malaria; artesunate; artemisinin; computer simulation; pharmacology; clinical; sequestra- tion; pharmacokinetics. Plasmodium falciparum is the malaria species responsible for the cells, iRBCs) to microvascular endothelium, a process known as largest number of deaths worldwide [1] and presents clinically sequestration. iRBC sequestration induces pathology through 3 in 2 forms. Patients with “uncomplicated” malaria have a rela- main causes: (1) impairing blood flow to organs through direct tively mild fever, are conscious, and capable of taking oral drug physical blockage of the capillaries [6], (2) indirect blockage via regimens; prompt treatment of uncomplicated malaria is associ- host defense mechanisms such as inflammation [3, 7], and (3) ated with low mortality [2]. Patients with “severe” malaria present physical damage to microvascular endothelium and the blood/ OA-CC-BY with 1, or a combination, of 4 syndromes: severe anemia, respira- brain barrier [8]. High case fatality rates occur, even if the drug tory distress, metabolic derangement, and cerebral malaria [3, 4]. kills parasites within sequestered iRBCs, because the molecules Patients are treated with parenteral artesunate, which rapidly kills responsible for sequestration (eg, P.  falciparum erythrocyte parasites, but resolution of pathology lags behind parasite killing; membrane protein 1 [9]) are still present on iRBC surfaces and case fatality rates are high even once patients have been admitted it takes a significant amount of time for these ligands to decline to the formal health system (typically between 5% and 12% [2] sufficiently for the sequestered iRBC to detach and/or for the although these have been falling to approximately 2% [5]). pathology associated with sequestration to resolve [10, 11]. A key factor responsible for severe malaria is the binding of Parasite clearance rates are a commonly used clinical outcome parasitized erythrocytes (subsequently called infected red blood measure to compare efficacy of antimalarial treatment regimens. However, parasite clearance rates correlate poorly with disease outcome in severe malaria. Large trials comparing intramuscu- Received 31 August 2018; editorial decision 3 October 2018; accepted 7 November 2018; published online December 5, 2018. lar artemether with quinine in African children showed more Present affiliation: Department of Molecular and Clinical Pharmacology, Institute of rapid parasite clearance with artemether but no difference Translational Medicine, University of Liverpool, United Kingdom. in case fatality [12, 13]. With parenteral artesunate, parasite Present affiliation: Metrum Research Group LLC, Tariffville, Connecticut. Correspondence: S.  Jones, MSc, Liverpool School of Tropical Medicine, Pembroke Place, clearance rates are not different in patients dying from severe Liverpool L3 5QA, United Kingdom (Sam.Jones@LSTMed.ac.uk). malaria compared to survivors (results cited in [14]). There are The Journal of Infectious Diseases 2019;219:1243–53 2 potential explanations why parasite clearance is an unsuitable © The Author(s) 2018. Published by Oxford University Press for the Infectious Diseases Society outcome measure in severe malaria: Firstly, parasite clearance of America. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted rates following treatment for uncomplicated malaria appear reuse, distribution, and reproduction in any medium, provided the original work is properly cited. to mainly reflect host immunity rather than drug effectiveness DOI: 10.1093/infdis/jiy649 Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1243 Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 [15–17] so may be a poor metric of overall drug effectiveness. response that may also contribute to pathology [3, 24]. It is unlikely Secondly, parasite clearance rates are measured on circulating that the iRBC immediately ruptures on death of the parasites parasites [15] whereas noncirculating, sequestered parasites are (which would reduce physical blockage of the capillary) or that the responsible for most clinical symptoms, pathology, and deaths immune/inflammatory responses immediately disappear when associated with severe malaria [3]. We developed a new model the parasite dies, so we assumed that pathology persists for a period based on existing pharmacokinetic/pharmacodynamic (PK/PD) aer ft the death of the sequestered parasites. We captured this effect models [18, 19] (themselves based on [20–22]) to investigate 2 by defining a “pathological recovery rate,” r, which is the rate at simple metrics reflecting the pathology of sequestered parasites which the pathology caused by sequestered iRBCs disappears with in severe malaria: the maximum sequestered load posttreat- time following the death of the parasite. As will be discussed later, ment, and the area under the curve (AUC) of sequestered para- there are no clinical estimates of this “recovery rate” so our strat- sites over time posttreatment. We quantified and compared the egy was to quantify the impact of dosing regimen and artemisinin impact of existing and proposed drug regimens on these metrics resistance across a range of values of recovery rate to test whether to identify rational drug dosing regimens for treatment of severe our results were dependent on assumed values for recovery rate malaria. Additionally, we quantified the likely impact of artemis- (we show later that they were not). We varied the “recovery rate” inin resistance in treatment of severe malaria. r in the simulations by altering its half-life (Table 1), which is the time it takes pathology caused by dead sequestered parasites to METHODS reduce by half. We assumed that parasite death, with consequent rupturing of the iRBC or reduction of binding ligands (allowing We utilized a computer-based PK/PD model to track changes iRBCs to detach from blood vessel walls), was essential to allow the in the number of sequestered iRBCs following drug admin- start of pathological recovery, hence sequestered iRBCs with liv- istration. The model was implemented in the statistical pro- ing parasites were not subject to the pathological recovery rate. We gramming software R [23] version 3.4.1. P. falciparum parasites quantified the pathological load L (t) at any time t posttreatment as undergo a 48-hour developmental cycle in human erythrocytes the sum of the current number of sequestered iRBCs with living with 2 main implications for pathology and treatment. Firstly, parasites α(t) and the lingering pathological effects of once-seques- parasites initially circulate freely in blood vessels but sequester tered iRBC whose parasites were killed in the current or previous (ie, bind to capillaries) at mature stages of their intraerythro- time periods, β(i), that is cytic cycle. Secondly, parasites differ in their sensitivity to drugs over the course of this 48-hour cycle. −− () ti r Lt () = αβ () ti + ()e (1) We assumed that severe malaria pathology is caused by a sin- i=1 gle clone (discussed in Supplementary Information) and simu- We used 2 metrics to analyze treatment regimens and resistance: lated a monoclonal infection. As previously described [22], we (1) maximum pathological load (MPL), the maximum value of separated the parasite population within a patient into 48 “age- L(t) occurring during a defined time period posttreatment, and bins” that each represent a 1-hour long development stage in (2) the area under the pathological load curve (AUC ) during a the parasite’s 48-hour life cycle within human erythrocytes. PL defined time period posttreatment, that is the total pathology in Parasites within age-bins have differing propensities to seques- that period. For example, the AUC in the period 0 to 24 hours ter and have varying degrees of drug sensitivity. Our model PL posttreatment is: tracked the number of iRBCs in each of 4 classes at any time posttreatment depending on whether the parasites are alive or AUCL = t () (2) PL t =1 dead, and whether the iRBC is circulating or sequestered: alive and circulating, alive and sequestered, dead and circulating, Simulating Patient Treatment Cohorts and dead and sequestered (see Figure  1 for illustration). Note We simulated a cohort of 10 000 patients who had parasitolog- that iRBCs classed as “dead and sequestered” are those iRBCs ical, pharmacological, and patient-specific parameters drawn whose parasites have died while sequestered and are either: (1) from the distributions given in Table 1. Individual patient proles fi still sequestered and causing pathology or (2) have ruptured/ allowed individual PK/PD variation to be incorporated to gen- detached from the capillary but are still associated with contin- erate individual patient posttreatment parasite clearance dynam- ued, lingering pathology. For model specification and details, ics (Supplementary Information). Each patient was simulated 3 see Supplementary Information. times under different scenarios: once for drug-sensitive parasites Pathological Load and Pathological Recovery Rate treated by the standard World Health Organization (WHO) reg- Severity of the malaria infection is determined by what we refer imen (2.4  mg/kg artesunate twice a day in the first 24 hours), to as “pathological load,” that is the number of sequestered iRBCs once for sensitive parasites treated with the simplified regimen (containing either living or dead parasites) physically restricting (4  mg/kg artesunate once a day, as proposed by Kremsner et  al blood flow and/or eliciting patient’s immune and/or inflammatory [30]), and once for artemisinin-resistant parasites treated by the 1244 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Parasites at any given time point Age post-invasion [h] 12 34 56 78 91011121313[...] 43 44 45 46 47 48 Proportion of parasites sequestering 00 00 00 00 00 00.2 0.40.5 ... 11 11 11 Parasite sensitivity to artesunate* 0 10 10 10 00.1 0.10.1 0.10.1 0.10.1 0.10.1 ... 10 1 00 0 *with artemisinin-sensitive parasites Pathology resolved by pathological recovery rate Sequestered, dead + lingering pathology Pathological load Sequestered, alive Multiply by PMR Circulating, alive Circulating, dead Parasites cleared by spleen 1e +12 1e +12 1e +09 1e +11 Alive, circulating parasites 1e +06 Dead circulating parasites Alive, sequestered parasites Pathological load 1e +10 1e +03 1e +00 1e +09 Time post treatment [h] Figure 1. A schematic of how our model tracks parasitemia and pathology posttreatment. A, How the simulation tracks parasitemia and pathology. The parasite population is separated into 48 hourly “age-bins” corresponding to their developmental age within their 48-hour intraerythrocytic cycle. A certain proportion of parasites in each age-bin will be sequestered, with 0% of parasites sequestering in age-bins 1 to 11 and approximately 100% sequestering in age-bins 14–48 (the proportions given in the figure are illustrative). Parasites in age-bin 48 rupture to produce new “daughter” parasites that enter age-bin 1; the number of daughter parasites that successfully invade new eryth- rocytes is the parasite multiplication rate (PMR). The simulation runs in 1-hour time steps and, if drug is present, it kills parasites according to their drug sensitivity, which is given in the second row of boxes as a proportion of basal kill rate (see Supplementary Information). Parasites that survive drug action are moved forward 1 age-bin (unless they are in age-bin 48 in which case they rupture to produce daughter parasites as described above). Parasites killed by drug in the time-step have 2 fates depending on their status. Those killed in circulating stages enter a pool of “dead circulating parasites” and will eventually be removed by splenic or other host clearance mechanisms. Those parasites that are killed while sequestered are removed from the simulation but their pathology does not instantly disappear with their death, so we track their the number of dead sequestered parasites and the lingering pathology of these parasites (second term of Equation 1) that resolves at the user-defined “pathological recovery rate”. B, How this methodology is used to simulate treatment of 1 exemplar individual (recall that patients and their parasites differ in a range of important variables; see Table 1). The number of alive circulating plus dead circulating parasites can be tracked over time posttreatment. These 2 classes can be directly observed (but not distinguished) in human blood samples and their rate of clearances, usually known as “parasite clearance rate” is often used as a proxy of clinical outcome; this enables us to verify that our simulations recovered these clinical observations. Live sequestered parasites are added to the lingering effects of sequestered parasites killed in earlier stages (ie, those contributing to “post-mortem pathology”) to obtain the pathological load L(t) at any time point posttreatment (Equation 1). Note that the number of dead sequestered parasites and their lingering pathology are not plotted here because that line is nearly indistinguishable from total pathological load (which includes live sequestered parasites) and so only total pathological load is plotted (right y axis; note difference in axis scale compared to other model compartments plotted on the left y axis). The dynamics of L(t) following treatment are used to calculate our key pathology metrics that are area under the pathology curve (AUC ) and the maximum parasite load (MPL). The patient dis- PL played in this figure had sensitive parasites and was treated with the standard regimen, with PK parameters drawn from Table 1. Abbreviation: iRBC, infected red blood cell. standard WHO regimen. This allowed us to compare the 2 dosing model parameters and dependent variables (ie, the pathology regimens (“standard” vs “simplified”) and the impact of resistance metrics AUC and MPL). PL (“sensitive” vs “resistant”) in each patient. Follow-up time was 48 All parameters are quantitative so can enter the PRCC without hours aer dr ft ug administration; this reflected a whole parasite modification. The exception is mean age-bin which, although life cycle within an iRBC but, more importantly, covers the period numeric, has a “circular” scale, age-bin 1 being adjacent to age- posttreatment where a patient is most likely to die [31, 32]. bin 48, due to parasites from ruptured iRBCs (at hour 48) rein- vading to restart the asexual life cycle. The mean age-bin variable Sensitivity Analysis was therefore split into either 5 or 3 ordinal classes (depending We conducted partial rank correlation coefficient (PRCC) using on whether parasites were hypersensitive or resistant to artemis- Spearman ρ to establish the strength of the relationship between inin), as described in Supplementary Information. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1245 Number of iRBC +1 Pathological load (parasite hours) Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Table 1. Parameter Values Used in the Simulations Parameter Unit Abbreviation Range Format Distribution Justification Initial parasite number P Double Uniform [25, 27] 10x, where () xx ∈< R |101 < 2 Mean of initial age-bin [h] Mean Integer Triangular with [25, 26], Supplementary distribution x + 0.5, where () xx ∈≤ N|0 ≤ 47 mode = 10 Information Standard deviation of initial [h] SD Integer Uniform [27], Supplementary age-bin distribution x, where () xx ∈≤ N|2 ≤ 4 Information Parasite multiplication rate PMR Integer Triangular with [26, 27] x, where () xx ∈≤ N|1 ≤ 10 mode = 1 −1 Pathological recovery rate [h ] r = ln(2)/x Integer Uniform half-life x, where () xx ∈≤ N|4 ≤ 12 −1 Splenic clearance rate [h ] u = ln(2)/x x, where mean = 2.7 and CV = 0.3 Double Normal [28, 29] half-life Half-maximum inhibitory [mg/L] IC50 x, where mean = 0.0016 and CV = 0.86 Double Log-normal [20] AS concentrations AS Half-maximum inhibitory [mg/L] IC50 x, where mean = 0.009 and CV = 1.17 Double Log-normal [20] DHA concentrations DHA −1 Maximal rate of drug killing [h ] V x, where mean = 1.78 and CV = 0.1 Double Normal [20, 22] max Slope factor n x, where mean = 4 and CV = 0.3 Double Normal [20] Abbreviations: AS, artesunate; CV, coefficient of variation; DHA, dihydroartemisinin. Not including volume of distribution (V ) / clearance (Cl). See Supplementary information for discussion of those parameters. e f Th ollowing parameters were included in the PRCC for example a ratio of 5 for resistant versus sensitive parasites analysis: indicates pathological metrics are 5 times higher when treating resistant parasites. We investigated 4 time periods posttreat- • duration of artesunate killing posttreatment; this captures all ment: 0–12 hours, 0–24 hours, 12–24 hours, and 24–48 hours. the PK/PD parameters in Table 1 except maximal artesunate kill rate Consistency of Model Outputs with Existing Field Data • maximal rate of artesunate killing (V ) Our model calculated parasite reduction ratios (PRR) from cir- max • initial mean age-bin as a categorical variable (see above) culating parasite numbers (Supplementary Information). The • variation of initial age-bin distribution (measured as the clinical endpoint of the trials by Kremsner and colleagues was the standard deviation (SD) around the mean) proportion of patients in each arm whose PRR at 24 hours (PRR ) • initial parasite number was >99% [30], reported as 79% and 78% for the 5-dose standard • parasite multiplication rate (PMR) and the 3-dose simplified regimen, respectively. When calibrated • half-life of the ‘pathological recovery rate’ (r). with PK parameters from Kremsner’s study [30], our results were consistent with these clinical observations, that is our model pre- e s Th plenic clearance rate was not included in the analysis as it dicted 78% and 74% for the standard and simplified regimen with has no impact on sequestered iRBC based pathology. hypersensitive parasites, respectively (Supplementary Table  3). However, the results we present below are calibrated using PK RESULTS parameters from Hendriksen et  al [33] (see Supplementary Information for justification), with which we observed lower val- Our model calculated pathological load and returns 2 outcome ues of 70% and 62% of patients with PRR >99% for the standard metrics: AUC and MPL. Figure  2 shows the values of these PL and simplified intramuscular regimens, respectively. metrics for 3 model scenarios: patients with sensitive parasites Hendriksen et al [33] do not report the percentage of patients treated with the standard WHO regimen, a comparison of the with PRR >99% in their study, so we could not simultane- ratios of AUC and MPL for treatment with simplified regimen PL ously compare the findings of our simulation with the find- versus standard regimen, and the impact or artemisinin resis- ings of Kremsner et al [30] and Hendriksen et al [33]. However, tance on outcomes following treatment with standard WHO Hendriksen et al [33] reported the population geometric mean of regimen. the fractional reduction in parasite counts at 24 hours as 96% (95% Ratios of outcome metrics are calculated as simplified reg- confidence interval [CI], 94%–98%,) following treatment with the imens scaled by standard regimen and as resistant parasites standard regimen. The population geometric mean obtained for scaled by sensitive parasites. High metrics are deleterious, thus the reduction in parasite counts at 24 hours (ie, PRR ) in our sim- ratios of >1 indicate worse prognosis associated with the sim- ulation using parameters from Hendriksen et al [33] was >99%. plified or resistant parasites. These ratios quantify the impact, 1246 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h 0-12h 0-24h 12-24h 24-48h 0-12h 0-24h 12-24h 24-48h A BC Sensitive parasites treated with standard regimenSimplified versus Standard regimenResistant versus Sensitive parasites 1e +15 4 2 1e +13 2 1 1e +11 1e +09 0 0 1e +15 4 2 1e +13 2 2 1e +11 1e +09 0 0 4 2 1e +15 1e +13 2 1 1e +11 1e +09 0 0 1e +15 4 2 1e +13 2 1 1e +11 1e +09 0 0 AUC MPL AUC MPL AUC MPL PL PL PL Outcome metric Outcome metric Outcome metric Figure 2. Values of area under the pathology curve (AUC ) and maximum parasite load (MPL) obtained for each of 3 model scenarios across 4 time periods posttreatment: PL 0–12 hours, 0–24 hours, 12–24 hours, and 24–48 hours. A, The “baseline scenario” when artemisinin-sensitive parasites are treated with the standard regimen. B, A com- parison of the simplified versus standard regimen (values >1 show the standard regimen is superior). C, A comparison of the standard regimen when used to treat resistant versus sensitive parasites (values >1 show that sensitive parasites produce better outcomes). e g Th eneral accepted value for PRR following artemisinin median ratio in 0–24 hours of 1.03; MPL was 1. At 24–48 hours, –4 treatment is 10 [34], which is very close to the value obtained higher medians of 1.49 and 1.45 for AUC and MPL, respec- PL here: for the standard regimen, using the artesunate  killing tively, were observed (Figure 2; Supplementary Table 4). duration derived from Hendriksen’s PK parameters (Figure  3) Parameter analysis with PRCC (Supplementary Table  8) −5 we obtained a mean PRR of 5.18 (see Supplementary revealed that patients whose initial infections were in either Information for a nuanced discussion of PK parameters). very late or very early initial mean age-bins (Figure  5, lower panel) will have worse outcomes with the simplified regimen. Standard Regimen Treatment of Artemisinin-Sensitive Parasites This occurred because parasites in these stages are largely insen- We simulated treatment of drug-sensitive parasites with the sitive to artesunate at first treatment, and the simplified regimen standard regimen and identified the key drivers of pathology by lacks the second dose, 12 hours later, of the standard regimen calculating which parameters were most correlated with AUC PL that would effectively target these parasites that had matured and MPL (Figure 4; Supplementary Table 7). The most highly into more artemisinin sensitive age-bins. correlated parameter for both metrics was the initial parasite e h Th alf-life of the recovery rate r had a moderate correlation number: large positive PRCCs (between 0.88 and 0.98) were with outputs in the 12 to 24-hour and 24 to 48-hour periods, observed with associated P values ≤.001 at all time periods. The indicating that assumption of slower recovery made the simpli- half-life of the recovery rate r had PRCC of 0.46 for AUC and PL fied regimen perform relatively better (Supplementary Figure 5). 0.34 for MPL in the 24 to 48-hour time period (P values ≤.001), We are confident this parameter does not aeff ct the valid- but PRCC of <0.3 in earlier time periods. All other parameters ity of our results; for complete discussion see Supplementary had PRCC values of <0.3, indicating that outcome metrics were Information. No other parameters have notable correlation not highly correlated as per accepted statistical criteria [35]. All with sequestration-based pathology when comparing regimens. other model parameters had negligible correlation. The most We repeated this analysis to compare regimens (ie, WHO likely explanation is that such a large proportion of parasites standard vs simplified) when treating artemisinin-resistant par - are killed by artesunate that small differences in the number asites. Differences between regimens were extremely similar to killed are negligible compared to the initial parasite number those shown in Figure  5 and are displayed in Supplementary and pathological recovery rate. Figure 7 and Supplementary Table 9. Comparison of Simplified and Standard Regimen The Impact of Artemisinin Resistance on Treatment by the Standard We evaluated alternative treatment regimens on artemisi- Regimen nin-sensitive parasites. These results are presented as ratios of Unsurprisingly, ratios of AUC and MPL when comparing PL AUC and MPL. The simplified regimen had a slightly higher resistant and sensitive parasites are never less than 1 (Figure 2), PL Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1247 Value of outcome metric Ratio of outcome metric Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 123456 7 Duration of artesunate killing with 2.4 mg/kg [h] 2 34567 Duration of artesunate killing with 4 mg/kg [h] Figure 3. Distribution of artesunate killing duration. Data for 10 000 patients following treatment with a single dose of artesunate of either 2.4 mg/kg (A) or 4 mg/kg (B); note the duration includes that of the active metabolite dihydroartemisinin. This distribution was obtained using parameters from Hendriksen et al [33]. that is under no circumstance did patients have a better out- posttreatment largely coincides with parasites in age-bins come when parasites are resistant. Differences in median values insensitive to artesunate through resistance, rendering the ini- (Figure 2; Supplementary Table 4) were extremely small. tial dose nearly or completely ineffective. We carried out PRRC analysis (Supplementary Table  10) to SD of the initial mean age-bin had a positive correlation with investigate whether this small difference obscured the presence the ratio (indicating that resistant parasites had worse outcomes of a vulnerable subgroup of patients. This appeared to be the as SD increased). This occurred because higher SD “nudged” case: patients whose infections are clustered in the early age- parts of the age-bin distribution into (or out of) resistant bins at time of treatment had pathological outcomes that were age-bins (ie, the contiguous bin 45–48 and 1–5 where killing significantly worse in the presence of resistance (Figure 6). is absent). PRCC analysis showed no other parameter had a In these early age-bins, ratios for AUC and MPL are as PRCC value of >0.01, suggesting the initial mean age-bin (and, PL high as 5 in the 0 to 24-hour period (comparisons based on the to a lesser extent, its SD) are the sole determinants of whether upper quartile value). This occurs because artesunate presence a patient’s outcome will be worse in the presence of resistance. 1248 • JID 2019:219 (15 April) • Jones et al Counts Counts Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 10.0 10.5 11.0 11.5 12.0 10.0 10.5 11.0 11.5 12.0 Initial parasite numbers 10^ Figure  4. Analysis of the baseline scenario. The impact of underlying factors on the standard World Health Organization regimen used to treat patients with artemis- inin-sensitive parasites. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on values of area under the pathology curve (AUC ) and PL maximum pathological load (MPL) obtained from a population. B, Values of AUC and MPL are plotted against the most highly correlated parameter, that is initial parasite PL number, for 4 time periods posttreatment. DISCUSSION was highly flexible (discussed in Supplementary Information) and, of necessity, reflected the limitations in our understanding We established a PK/PD modeling methodology capable of of pathology, for example how rapidly pathology is resolved fol- investigating the treatment of severe malaria. Kremsner et  al lowing parasite death and whether pathology depends on max- [36] recognized the clinical necessity of this, and noted that imal sequestered load (measured as MPL) or on total exposure “for the first time, we [ie, Kremsner et al] are assessing artesu- (measured as AUC ). An interesting, highly important result is PL nate using similar pharmacokinetic and dynamic approaches”. that the key quantitative assumption made in the analysis, the Parasite clearance is likely to be a poor measure of regimen rate of resolution of pathology (measured as the half-life of r), had effectiveness (and, by extension, clinical outcome) in severe little effect on our conclusions when comparing alternative regi- malaria where pathology is due to sequestered parasites. The mens or the impact of resistance (Supplementary Information) effects of alternative regimens and the impact of drug resistance implying that the pathological model is a robust to assumptions can only be investigated by traditional clinical outcomes using made in this comparative investigation. Importantly, while cir- large-scale clinical trials, so pharmacological modeling of the culating parasite loads do not reflect the pathology of severe type proposed here is essential to help generate the evidence malaria they are currently the regular endpoint of choice in base for rational treatment design. Our pathological modeling Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1249 Value of outcome metric PRCC Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 110203040481 10 20 30 40 48 Mean age bin distribution [h] Figure 5. Evaluation of alternative drug treatment regimens. Comparison of the simplified versus World Health Organization standard regimen for treatment of artemis- inin-sensitive parasites; ratios of >1 indicate the simplified regimen produces worse outcome metrics. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on the ratios of area under the pathology curve (AUC ) and maximum pathological load (MPL). B, Ratios of AUC and MPL are plotted against the most PL PL highly correlated parameter (initial mean age-bin), for 4 time periods posttreatment. severe malaria trials, including those undertaken by Kremsner lower median AUC within the first 24 hours posttreatment PL et al [30, 37]; our model was able to reproduce the clinical out- (Figure  2; Supplementary Table 4). This difference was greater comes reported in [30, 33] (when appropriately parameterized), in the 24 to 48-hour period, but the majority of pathological and recover expected PRR , so we are confident it is reflective load occurred within the first 24 hours as artesunate rapidly of in vivo scenarios (Supplementary Information). kills parasites: AUC in the 24 to 48-hour period is, on aver- PL Kremsner and colleagues [30, 37] concluded that their sim- age, between 20% and 30% that of AUC in the 0 to 24-hour PL plified regimen was noninferior to the standard WHO regimen period (data not shown). The first 24 hours are critical for and possessed operational advantages due to less frequent drug patient survival [31], so outcome metrics at 24–48 hours may administration [30, 37]. This work was influential and ini- have little relevance in choosing between regimens. However, tiated a wider debate about the best drug regimen(s) to treat the simplified regimen performed much worse in the subgroup severe malaria [14, 36, 38] to which our study can contribute. of patients with very late or very early initial mean age-bins. Comparison of the 0 to 24-hour and 12 to 24-hour period was Based on these results, we are dubious about recommending used to compare the effects of the initial, larger dose of the sim- use of the simplified regimen but add an important rider to this. plified regimen against the additional dose at 12 hours with Kremsner et al never claimed this simplified regimen would be the standard regimen. The standard regimen produced slightly superior, but argued that any inferiority, if it exists, would be 1250 • JID 2019:219 (15 April) • Jones et al PRCC Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 110203040481 10 20 30 40 48 Mean age bin distribution [h] Figure 6. Analysis of the impact of artemisinin resistance. The effectiveness of the World Health Organization standard regimen used to treat resistant versus sensitive parasites; ratios of >1 indicate that resistant parasites have worse outcome metrics. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on the ratios of area under the pathology curve (AUC ) and maximum pathological load (MPL). B, Ratios of AUC and MPL are plotted against the most highly correlated PL PL parameter (mean age-bin), for 4 time periods posttreatment. within acceptable margins. We leave it to clinically qualified malaria (provided there was no resistance to partner drugs) personnel to judge whether 50% in some subgroups is within an but artemisinin resistance clearly poses a much larger threat acceptable margin of inferiority, especially given our inability to treatment of severe malaria than it does to uncomplicated to directly link our pathological outcomes with the likelihood malaria. Although differences between sensitive and resistant of mortality. parasites across the entire population are minor (Figure  2; We assessed the impact of artemisinin resistance on treat- Supplementary Table 4), there is an extremely vulnerable sub- ment of severe malaria, that is the extent to which resistance group of patients whose infections at the time of treatment are increased MPL and AUC . Resistance prevents drug killing in clustered in very late or very early age-bins (ie, where parasites PL age-bins 2–4 (these bins are otherwise hypersensitive) resulting are resistant in our model; Figure 6). in no killing for a contiguous 8-hour period in resistant par- Note that we specifically model relatively tightly synchro- asites (ie, age-bins 45 to 5). Our results show the initial mean nized parasite distributions (Supplementary Information); if age-bin and its SD are the only parameters that distinguish out- distributions were to become less tightly synchronized the vul- comes between sensitive and resistance parasites (Figure 6). We nerability of patients with early initial mean age-bins decreases argued previously [39] that artemisinin resistance would have and both the difference between regimens and the impact of a negligible impact on eventual cure rates in uncomplicated resistance reduces. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1251 PRCC Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 We present a highly adaptable methodology for PK/PD mod- 6. Dondorp AM, Kager PA, Vreeken J, White NJ. Abnormal eling of treatment of severe malaria that was able to recover key blood flow and red blood cell deformability in severe clinical observations (based on circulating parasite numbers), malaria. Parasitol Today 2000; 16:228–32. and, with novel metrics, used to investigate the pathology of 7. Clark IA, Alleva LM. Is human malarial coma caused, or severe malaria. Our model showed that while on a popula- merely deepened, by sequestration? Trends Parasitol 2009; tion level a simplified artesunate regimen is noninferior to the 25:314–8. standard WHO regimen, outcomes in a subgroup of patients 8. Medana IM, Turner GD. Plasmodium falciparum and the with infections grouped in late or early initial mean age-bins blood-brain barrier–contacts and consequences. J Infect are notably worse with the simplified regimen. The emergence Dis 2007; 195:921–3. of artemisinin resistance in early ring stages poses a significant 9. Smith JD, Rowe JA, Higgins MK, Lavstsen T. Malaria’s threat to this same group of patients. Neither of these results are deadly grip: cytoadhesion of Plasmodium falciparum-in- particularly obvious from summary statistics of the population fected erythrocytes. Cell Microbiol 2013; 15:1976–83. and so subgroup analysis is particularly important in devising 10. Hughes KR, Biagini GA, Craig AG. Continued cytoadherence treatment strategies for severe malaria. of Plasmodium falciparum infected red blood cells aer a ft nti- malarial treatment. Mol Biochem Parasitol 2010; 169:71–8. Supplementary Data 11. Udomsangpetch R, Pipitaporn B, Krishna S, et  al. Supplementary materials are available at e Th Journal of Infectious Antimalarial drugs reduce cytoadherence and rosetting Diseases online. Consisting of data provided by the authors to Plasmodium falciparum. J Infect Dis 1996; 173:691–8. benefit the reader, the posted materials are not copyedited and 12. van Hensbroek MB, Onyiorah E, Jaffar S, et  al. A trial of are the sole responsibility of the authors, so questions or com- artemether or quinine in children with cerebral malaria. N ments should be addressed to the corresponding author. Engl J Med 1996; 335:69–75. 13. Esu E, Effa EE, Opie ON, Uwaoma A, Meremikwu MM. Notes Artemether for severe malaria. Cochrane Database Syst Rev Acknowledgment. We thank 2 anonymous reviewers for 2014; (9):CD010678. their helpful comments and critique. 14. Dondorp AM, Maude RJ, Hendriksen IC, Day NP, White Financial support. This work was supported by the UK NJ. Artesunate dosing in severe falciparum malaria. J Infect Medical Research Council (grant numbers G1100522 and MR/ Dis 2012; 206:618–9; author reply 622. L022508/1); the Bill and Melinda Gates Foundation (grant 15. Hastings IM, Kay K, Hodel EM. How robust are malaria par- number1032350); and the Malaria Modeling Consortium asite clearance rates as indicators of drug effectiveness and (grant number UWSC9757). resistance? Antimicrob Agents Chemother 2015; 59:6428–36. Potential coni fl cts of interest. All authors: No reported con- 16. Ferreira PE, Culleton R, Gil JP, Meshnick SR. Artemisinin flicts of interest. All authors have submitted the ICMJE Form resistance in Plasmodium falciparum: what is it really? for Disclosure of Potential Conflicts of Interest. Conflicts that Trends Parasitol 2013; 29:318–20. the editors consider relevant to the content of the manuscript 17. Krishna S, Kremsner PG. Antidogmatic approaches to arte- have been disclosed. misinin resistance: reappraisal as treatment failure with arte- misinin combination therapy. Trends Parasitol 2013; 29:313–7. References 18. Hastings IM, Hodel EM, Kay K. Quantifying the pharma- cology of antimalarial drug combination therapy. Sci Rep 1. World Health Organization (WHO). World malaria report 2016; 6:32762. 2016. Geneva, Switzerland: WHO, 2016. 19. Hodel EM, Kay K, Hastings IM. Incorporating stage-spe- 2. Olliaro P. Editorial commentary: mortality associated with cific drug action into pharmacological modeling of antima- severe Plasmodium falciparum malaria increases with age. larial drug treatment. Antimicrob Agents Chemother 2016; Clin Infect Dis 2008; 47:158–60. 60:2747–56. 3. Cunnington AJ, Riley EM, Walther M. Stuck in a rut? 20. Kay K, Hastings IM. Improving pharmacokinetic-phar- Reconsidering the role of parasite sequestration in severe macodynamic modeling to investigate anti-infective che- malaria syndromes. Trends Parasitol 2013; 29:585–92. motherapy with application to the current generation of 4. Mackintosh CL, Beeson JG, Marsh K. Clinical features antimalarial drugs. PLoS Comput Biol 2013; 9:e1003151. and pathogenesis of severe malaria. Trends Parasitol 2004; 21. Kay K, Hodel EM, Hastings IM. Altering antimalarial drug 20:597–603. regimens may dramatically enhance and restore drug effec- 5. Taylor T, Olola C, Valim C, et al. Standardized data collec- tiveness. Antimicrob Agents Chemother 2015; 59:6419–27. tion for multi-center clinical studies of severe malaria in 22. Hodel EM, Kay K, Hastings IM. Incorporating stage-spe- African children: establishing the SMAC network. Trans R cific drug action into pharmacological modeling of Soc Trop Med Hyg 2006; 100:615–22. 1252 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 antimalarial drug treatment. Antimicrob Agents Chemother 31. Kendjo E, Agbenyega T, Bojang K, et al. Mortality patterns 2016; 60:2747–56. and site heterogeneity of severe malaria in African children. 23. R Core Team. R: A language and environment for statisti- PLoS One 2013; 8:e58686. cal computing. Vienna, Austria: R Foundation for Statistical 32. Maitland K. Management of severe paediatric malaria in Computing, 2014. resource-limited settings. BMC Med 2015; 13:42. 24. Chua CL, Brown G, Hamilton JA, Rogerson S, Boeuf P. 33. Hendriksen IC, Mtove G, Kent A, et al. Population pharma- Monocytes and macrophages in malaria: protection or cokinetics of intramuscular artesunate in African children pathology? Trends Parasitol 2013; 29:26–34. with severe malaria: implications for a practical dosing reg- 25. White NJ. The parasite clearance curve. Malar J 2011; imen. Clin Pharmacol Ther 2013; 93:443–50. 10:278. 34. White NJ. Assessment of the pharmacodynamic prop- 26. White NJ. Malaria parasite clearance. Malar J 2017; 16:88. erties of antimalarial drugs in vivo. Antimicrob Agents 27. Saralamba S, Pan-Ngum W, Maude RJ, et al. Intrahost mod- Chemother 1997; 41:1413–22. eling of artemisinin resistance in Plasmodium falciparum. 35. Hinkle DE, Wiersma W, Jurs SG. Applied statistics for the Proc Natl Acad Sci U S A 2011; 108:397–402. behavioral sciences. 5th ed. Boston: Houghton Mifflin, 2003. 28. Gordi T, Xie R, Jusko WJ. Semi-mechanistic pharmaco- 36. Kremsner PG, Taylor T, Issifou S, et al. A simplified intrave- kinetic/pharmacodynamic modelling of the antimalarial nous artesunate regimen for severe malaria–reply. J Infect effect of artemisinin. Br J Clin Pharmacol 2005; 60:594–604. Dis 2012; 206:622. 29. Hietala SF, Martensson A, Ngasala B, et  al. Population 37. Kremsner PG, Taylor T, Issifou S, et al. A simplified intra- pharmacokinetics and pharmacodynamics of artemether venous artesunate regimen for severe malaria. J Infect Dis and lumefantrine during combination treatment in chil- 2012; 205:312–9. dren with uncomplicated falciparum malaria in Tanzania. 38. Woodrow CJ, Taylor WR. Reappraisal of the efficacy of a Antimicrob Agents Chemother 2010; 54:4780–8. simplified artesunate regimen in falciparum malaria. J 30. Kremsner PG, Adegnika AA, Hounkpatin AB, et  al. Infect Dis 2012; 206:619–21. Intramuscular artesunate for severe malaria in African chil- 39. Hastings IM, Hodel EM, Kay K. Quantifying the pharma- dren: a multicenter randomized controlled trial. PLoS Med cology of antimalarial drug combination therapy. Sci Rep 2016; 13:e1001938. 2016; 6:32762. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1253 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journal of Infectious Diseases Oxford University Press

Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance

Loading next page...
 
/lp/oxford-university-press/optimal-treatments-for-severe-malaria-and-the-threat-posed-by-IaY8EOGJ85

References (37)

Publisher
Oxford University Press
Copyright
Copyright © 2022 Infectious Diseases Society of America
ISSN
0022-1899
eISSN
1537-6613
DOI
10.1093/infdis/jiy649
Publisher site
See Article on Publisher Site

Abstract

Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance Sam Jones, Eva Maria Hodel, Raman Sharma, Katherine Kay, Ian M Hastings Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 The Journal of Infectious Diseases MAJOR ARTICLE Optimal Treatments for Severe Malaria and the Threat Posed by Artemisinin Resistance 1, 1,2,a 1 1,b 1 Sam Jones, Eva Maria Hodel, Raman Sharma, Katherine Kay, and Ian M. Hastings 1 2 Parasitology Department and Department of Clinical Sciences, Liverpool School of Tropical Medicine, United Kingdom (See the Editorial Commentary by Small and Seydel on pages 1176–7.) Background. Standard treatment for severe malaria is with artesunate; patient survival in the 24 hours immediately posttreat- ment is the key objective. Clinical trials use clearance rates of circulating parasites as their clinical outcome, but the pathology of XX severe malaria is attributed primarily to noncirculating, sequestered, parasites, so there is a disconnect between existing clinical metrics and objectives. XXXX Methods. We extend existing pharmacokinetic/pharmacodynamic modeling methods to simulate the treatment of 10 000 patients with severe malaria and track the pathology caused by sequestered parasites. Results. Our model recovered the clinical outcomes of existing studies (based on circulating parasites) and showed a “sim- plified” artesunate regimen was noninferior to the existing World Health Organization regimen across the patient population but resulted in worse outcomes in a subgroup of patients with infections clustered in early stages of the parasite life cycle. This same group of patients were extremely vulnerable to resistance emerging in parasite early ring stages. Conclusions. We quantify patient outcomes in a manner appropriate for severe malaria with a flexible framework that allows future researchers to implement their beliefs about underlying pathology. We highlight with some urgency the threat posed to treat- ment of severe malaria by artemisinin resistance in parasite early ring stages. Keywords. Plasmodium falciparum; malaria; artesunate; artemisinin; computer simulation; pharmacology; clinical; sequestra- tion; pharmacokinetics. Plasmodium falciparum is the malaria species responsible for the cells, iRBCs) to microvascular endothelium, a process known as largest number of deaths worldwide [1] and presents clinically sequestration. iRBC sequestration induces pathology through 3 in 2 forms. Patients with “uncomplicated” malaria have a rela- main causes: (1) impairing blood flow to organs through direct tively mild fever, are conscious, and capable of taking oral drug physical blockage of the capillaries [6], (2) indirect blockage via regimens; prompt treatment of uncomplicated malaria is associ- host defense mechanisms such as inflammation [3, 7], and (3) ated with low mortality [2]. Patients with “severe” malaria present physical damage to microvascular endothelium and the blood/ OA-CC-BY with 1, or a combination, of 4 syndromes: severe anemia, respira- brain barrier [8]. High case fatality rates occur, even if the drug tory distress, metabolic derangement, and cerebral malaria [3, 4]. kills parasites within sequestered iRBCs, because the molecules Patients are treated with parenteral artesunate, which rapidly kills responsible for sequestration (eg, P.  falciparum erythrocyte parasites, but resolution of pathology lags behind parasite killing; membrane protein 1 [9]) are still present on iRBC surfaces and case fatality rates are high even once patients have been admitted it takes a significant amount of time for these ligands to decline to the formal health system (typically between 5% and 12% [2] sufficiently for the sequestered iRBC to detach and/or for the although these have been falling to approximately 2% [5]). pathology associated with sequestration to resolve [10, 11]. A key factor responsible for severe malaria is the binding of Parasite clearance rates are a commonly used clinical outcome parasitized erythrocytes (subsequently called infected red blood measure to compare efficacy of antimalarial treatment regimens. However, parasite clearance rates correlate poorly with disease outcome in severe malaria. Large trials comparing intramuscu- Received 31 August 2018; editorial decision 3 October 2018; accepted 7 November 2018; published online December 5, 2018. lar artemether with quinine in African children showed more Present affiliation: Department of Molecular and Clinical Pharmacology, Institute of rapid parasite clearance with artemether but no difference Translational Medicine, University of Liverpool, United Kingdom. in case fatality [12, 13]. With parenteral artesunate, parasite Present affiliation: Metrum Research Group LLC, Tariffville, Connecticut. Correspondence: S.  Jones, MSc, Liverpool School of Tropical Medicine, Pembroke Place, clearance rates are not different in patients dying from severe Liverpool L3 5QA, United Kingdom (Sam.Jones@LSTMed.ac.uk). malaria compared to survivors (results cited in [14]). There are The Journal of Infectious Diseases 2019;219:1243–53 2 potential explanations why parasite clearance is an unsuitable © The Author(s) 2018. Published by Oxford University Press for the Infectious Diseases Society outcome measure in severe malaria: Firstly, parasite clearance of America. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted rates following treatment for uncomplicated malaria appear reuse, distribution, and reproduction in any medium, provided the original work is properly cited. to mainly reflect host immunity rather than drug effectiveness DOI: 10.1093/infdis/jiy649 Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1243 Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 [15–17] so may be a poor metric of overall drug effectiveness. response that may also contribute to pathology [3, 24]. It is unlikely Secondly, parasite clearance rates are measured on circulating that the iRBC immediately ruptures on death of the parasites parasites [15] whereas noncirculating, sequestered parasites are (which would reduce physical blockage of the capillary) or that the responsible for most clinical symptoms, pathology, and deaths immune/inflammatory responses immediately disappear when associated with severe malaria [3]. We developed a new model the parasite dies, so we assumed that pathology persists for a period based on existing pharmacokinetic/pharmacodynamic (PK/PD) aer ft the death of the sequestered parasites. We captured this effect models [18, 19] (themselves based on [20–22]) to investigate 2 by defining a “pathological recovery rate,” r, which is the rate at simple metrics reflecting the pathology of sequestered parasites which the pathology caused by sequestered iRBCs disappears with in severe malaria: the maximum sequestered load posttreat- time following the death of the parasite. As will be discussed later, ment, and the area under the curve (AUC) of sequestered para- there are no clinical estimates of this “recovery rate” so our strat- sites over time posttreatment. We quantified and compared the egy was to quantify the impact of dosing regimen and artemisinin impact of existing and proposed drug regimens on these metrics resistance across a range of values of recovery rate to test whether to identify rational drug dosing regimens for treatment of severe our results were dependent on assumed values for recovery rate malaria. Additionally, we quantified the likely impact of artemis- (we show later that they were not). We varied the “recovery rate” inin resistance in treatment of severe malaria. r in the simulations by altering its half-life (Table 1), which is the time it takes pathology caused by dead sequestered parasites to METHODS reduce by half. We assumed that parasite death, with consequent rupturing of the iRBC or reduction of binding ligands (allowing We utilized a computer-based PK/PD model to track changes iRBCs to detach from blood vessel walls), was essential to allow the in the number of sequestered iRBCs following drug admin- start of pathological recovery, hence sequestered iRBCs with liv- istration. The model was implemented in the statistical pro- ing parasites were not subject to the pathological recovery rate. We gramming software R [23] version 3.4.1. P. falciparum parasites quantified the pathological load L (t) at any time t posttreatment as undergo a 48-hour developmental cycle in human erythrocytes the sum of the current number of sequestered iRBCs with living with 2 main implications for pathology and treatment. Firstly, parasites α(t) and the lingering pathological effects of once-seques- parasites initially circulate freely in blood vessels but sequester tered iRBC whose parasites were killed in the current or previous (ie, bind to capillaries) at mature stages of their intraerythro- time periods, β(i), that is cytic cycle. Secondly, parasites differ in their sensitivity to drugs over the course of this 48-hour cycle. −− () ti r Lt () = αβ () ti + ()e (1) We assumed that severe malaria pathology is caused by a sin- i=1 gle clone (discussed in Supplementary Information) and simu- We used 2 metrics to analyze treatment regimens and resistance: lated a monoclonal infection. As previously described [22], we (1) maximum pathological load (MPL), the maximum value of separated the parasite population within a patient into 48 “age- L(t) occurring during a defined time period posttreatment, and bins” that each represent a 1-hour long development stage in (2) the area under the pathological load curve (AUC ) during a the parasite’s 48-hour life cycle within human erythrocytes. PL defined time period posttreatment, that is the total pathology in Parasites within age-bins have differing propensities to seques- that period. For example, the AUC in the period 0 to 24 hours ter and have varying degrees of drug sensitivity. Our model PL posttreatment is: tracked the number of iRBCs in each of 4 classes at any time posttreatment depending on whether the parasites are alive or AUCL = t () (2) PL t =1 dead, and whether the iRBC is circulating or sequestered: alive and circulating, alive and sequestered, dead and circulating, Simulating Patient Treatment Cohorts and dead and sequestered (see Figure  1 for illustration). Note We simulated a cohort of 10 000 patients who had parasitolog- that iRBCs classed as “dead and sequestered” are those iRBCs ical, pharmacological, and patient-specific parameters drawn whose parasites have died while sequestered and are either: (1) from the distributions given in Table 1. Individual patient proles fi still sequestered and causing pathology or (2) have ruptured/ allowed individual PK/PD variation to be incorporated to gen- detached from the capillary but are still associated with contin- erate individual patient posttreatment parasite clearance dynam- ued, lingering pathology. For model specification and details, ics (Supplementary Information). Each patient was simulated 3 see Supplementary Information. times under different scenarios: once for drug-sensitive parasites Pathological Load and Pathological Recovery Rate treated by the standard World Health Organization (WHO) reg- Severity of the malaria infection is determined by what we refer imen (2.4  mg/kg artesunate twice a day in the first 24 hours), to as “pathological load,” that is the number of sequestered iRBCs once for sensitive parasites treated with the simplified regimen (containing either living or dead parasites) physically restricting (4  mg/kg artesunate once a day, as proposed by Kremsner et  al blood flow and/or eliciting patient’s immune and/or inflammatory [30]), and once for artemisinin-resistant parasites treated by the 1244 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Parasites at any given time point Age post-invasion [h] 12 34 56 78 91011121313[...] 43 44 45 46 47 48 Proportion of parasites sequestering 00 00 00 00 00 00.2 0.40.5 ... 11 11 11 Parasite sensitivity to artesunate* 0 10 10 10 00.1 0.10.1 0.10.1 0.10.1 0.10.1 ... 10 1 00 0 *with artemisinin-sensitive parasites Pathology resolved by pathological recovery rate Sequestered, dead + lingering pathology Pathological load Sequestered, alive Multiply by PMR Circulating, alive Circulating, dead Parasites cleared by spleen 1e +12 1e +12 1e +09 1e +11 Alive, circulating parasites 1e +06 Dead circulating parasites Alive, sequestered parasites Pathological load 1e +10 1e +03 1e +00 1e +09 Time post treatment [h] Figure 1. A schematic of how our model tracks parasitemia and pathology posttreatment. A, How the simulation tracks parasitemia and pathology. The parasite population is separated into 48 hourly “age-bins” corresponding to their developmental age within their 48-hour intraerythrocytic cycle. A certain proportion of parasites in each age-bin will be sequestered, with 0% of parasites sequestering in age-bins 1 to 11 and approximately 100% sequestering in age-bins 14–48 (the proportions given in the figure are illustrative). Parasites in age-bin 48 rupture to produce new “daughter” parasites that enter age-bin 1; the number of daughter parasites that successfully invade new eryth- rocytes is the parasite multiplication rate (PMR). The simulation runs in 1-hour time steps and, if drug is present, it kills parasites according to their drug sensitivity, which is given in the second row of boxes as a proportion of basal kill rate (see Supplementary Information). Parasites that survive drug action are moved forward 1 age-bin (unless they are in age-bin 48 in which case they rupture to produce daughter parasites as described above). Parasites killed by drug in the time-step have 2 fates depending on their status. Those killed in circulating stages enter a pool of “dead circulating parasites” and will eventually be removed by splenic or other host clearance mechanisms. Those parasites that are killed while sequestered are removed from the simulation but their pathology does not instantly disappear with their death, so we track their the number of dead sequestered parasites and the lingering pathology of these parasites (second term of Equation 1) that resolves at the user-defined “pathological recovery rate”. B, How this methodology is used to simulate treatment of 1 exemplar individual (recall that patients and their parasites differ in a range of important variables; see Table 1). The number of alive circulating plus dead circulating parasites can be tracked over time posttreatment. These 2 classes can be directly observed (but not distinguished) in human blood samples and their rate of clearances, usually known as “parasite clearance rate” is often used as a proxy of clinical outcome; this enables us to verify that our simulations recovered these clinical observations. Live sequestered parasites are added to the lingering effects of sequestered parasites killed in earlier stages (ie, those contributing to “post-mortem pathology”) to obtain the pathological load L(t) at any time point posttreatment (Equation 1). Note that the number of dead sequestered parasites and their lingering pathology are not plotted here because that line is nearly indistinguishable from total pathological load (which includes live sequestered parasites) and so only total pathological load is plotted (right y axis; note difference in axis scale compared to other model compartments plotted on the left y axis). The dynamics of L(t) following treatment are used to calculate our key pathology metrics that are area under the pathology curve (AUC ) and the maximum parasite load (MPL). The patient dis- PL played in this figure had sensitive parasites and was treated with the standard regimen, with PK parameters drawn from Table 1. Abbreviation: iRBC, infected red blood cell. standard WHO regimen. This allowed us to compare the 2 dosing model parameters and dependent variables (ie, the pathology regimens (“standard” vs “simplified”) and the impact of resistance metrics AUC and MPL). PL (“sensitive” vs “resistant”) in each patient. Follow-up time was 48 All parameters are quantitative so can enter the PRCC without hours aer dr ft ug administration; this reflected a whole parasite modification. The exception is mean age-bin which, although life cycle within an iRBC but, more importantly, covers the period numeric, has a “circular” scale, age-bin 1 being adjacent to age- posttreatment where a patient is most likely to die [31, 32]. bin 48, due to parasites from ruptured iRBCs (at hour 48) rein- vading to restart the asexual life cycle. The mean age-bin variable Sensitivity Analysis was therefore split into either 5 or 3 ordinal classes (depending We conducted partial rank correlation coefficient (PRCC) using on whether parasites were hypersensitive or resistant to artemis- Spearman ρ to establish the strength of the relationship between inin), as described in Supplementary Information. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1245 Number of iRBC +1 Pathological load (parasite hours) Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 Table 1. Parameter Values Used in the Simulations Parameter Unit Abbreviation Range Format Distribution Justification Initial parasite number P Double Uniform [25, 27] 10x, where () xx ∈< R |101 < 2 Mean of initial age-bin [h] Mean Integer Triangular with [25, 26], Supplementary distribution x + 0.5, where () xx ∈≤ N|0 ≤ 47 mode = 10 Information Standard deviation of initial [h] SD Integer Uniform [27], Supplementary age-bin distribution x, where () xx ∈≤ N|2 ≤ 4 Information Parasite multiplication rate PMR Integer Triangular with [26, 27] x, where () xx ∈≤ N|1 ≤ 10 mode = 1 −1 Pathological recovery rate [h ] r = ln(2)/x Integer Uniform half-life x, where () xx ∈≤ N|4 ≤ 12 −1 Splenic clearance rate [h ] u = ln(2)/x x, where mean = 2.7 and CV = 0.3 Double Normal [28, 29] half-life Half-maximum inhibitory [mg/L] IC50 x, where mean = 0.0016 and CV = 0.86 Double Log-normal [20] AS concentrations AS Half-maximum inhibitory [mg/L] IC50 x, where mean = 0.009 and CV = 1.17 Double Log-normal [20] DHA concentrations DHA −1 Maximal rate of drug killing [h ] V x, where mean = 1.78 and CV = 0.1 Double Normal [20, 22] max Slope factor n x, where mean = 4 and CV = 0.3 Double Normal [20] Abbreviations: AS, artesunate; CV, coefficient of variation; DHA, dihydroartemisinin. Not including volume of distribution (V ) / clearance (Cl). See Supplementary information for discussion of those parameters. e f Th ollowing parameters were included in the PRCC for example a ratio of 5 for resistant versus sensitive parasites analysis: indicates pathological metrics are 5 times higher when treating resistant parasites. We investigated 4 time periods posttreat- • duration of artesunate killing posttreatment; this captures all ment: 0–12 hours, 0–24 hours, 12–24 hours, and 24–48 hours. the PK/PD parameters in Table 1 except maximal artesunate kill rate Consistency of Model Outputs with Existing Field Data • maximal rate of artesunate killing (V ) Our model calculated parasite reduction ratios (PRR) from cir- max • initial mean age-bin as a categorical variable (see above) culating parasite numbers (Supplementary Information). The • variation of initial age-bin distribution (measured as the clinical endpoint of the trials by Kremsner and colleagues was the standard deviation (SD) around the mean) proportion of patients in each arm whose PRR at 24 hours (PRR ) • initial parasite number was >99% [30], reported as 79% and 78% for the 5-dose standard • parasite multiplication rate (PMR) and the 3-dose simplified regimen, respectively. When calibrated • half-life of the ‘pathological recovery rate’ (r). with PK parameters from Kremsner’s study [30], our results were consistent with these clinical observations, that is our model pre- e s Th plenic clearance rate was not included in the analysis as it dicted 78% and 74% for the standard and simplified regimen with has no impact on sequestered iRBC based pathology. hypersensitive parasites, respectively (Supplementary Table  3). However, the results we present below are calibrated using PK RESULTS parameters from Hendriksen et  al [33] (see Supplementary Information for justification), with which we observed lower val- Our model calculated pathological load and returns 2 outcome ues of 70% and 62% of patients with PRR >99% for the standard metrics: AUC and MPL. Figure  2 shows the values of these PL and simplified intramuscular regimens, respectively. metrics for 3 model scenarios: patients with sensitive parasites Hendriksen et al [33] do not report the percentage of patients treated with the standard WHO regimen, a comparison of the with PRR >99% in their study, so we could not simultane- ratios of AUC and MPL for treatment with simplified regimen PL ously compare the findings of our simulation with the find- versus standard regimen, and the impact or artemisinin resis- ings of Kremsner et al [30] and Hendriksen et al [33]. However, tance on outcomes following treatment with standard WHO Hendriksen et al [33] reported the population geometric mean of regimen. the fractional reduction in parasite counts at 24 hours as 96% (95% Ratios of outcome metrics are calculated as simplified reg- confidence interval [CI], 94%–98%,) following treatment with the imens scaled by standard regimen and as resistant parasites standard regimen. The population geometric mean obtained for scaled by sensitive parasites. High metrics are deleterious, thus the reduction in parasite counts at 24 hours (ie, PRR ) in our sim- ratios of >1 indicate worse prognosis associated with the sim- ulation using parameters from Hendriksen et al [33] was >99%. plified or resistant parasites. These ratios quantify the impact, 1246 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h 0-12h 0-24h 12-24h 24-48h 0-12h 0-24h 12-24h 24-48h A BC Sensitive parasites treated with standard regimenSimplified versus Standard regimenResistant versus Sensitive parasites 1e +15 4 2 1e +13 2 1 1e +11 1e +09 0 0 1e +15 4 2 1e +13 2 2 1e +11 1e +09 0 0 4 2 1e +15 1e +13 2 1 1e +11 1e +09 0 0 1e +15 4 2 1e +13 2 1 1e +11 1e +09 0 0 AUC MPL AUC MPL AUC MPL PL PL PL Outcome metric Outcome metric Outcome metric Figure 2. Values of area under the pathology curve (AUC ) and maximum parasite load (MPL) obtained for each of 3 model scenarios across 4 time periods posttreatment: PL 0–12 hours, 0–24 hours, 12–24 hours, and 24–48 hours. A, The “baseline scenario” when artemisinin-sensitive parasites are treated with the standard regimen. B, A com- parison of the simplified versus standard regimen (values >1 show the standard regimen is superior). C, A comparison of the standard regimen when used to treat resistant versus sensitive parasites (values >1 show that sensitive parasites produce better outcomes). e g Th eneral accepted value for PRR following artemisinin median ratio in 0–24 hours of 1.03; MPL was 1. At 24–48 hours, –4 treatment is 10 [34], which is very close to the value obtained higher medians of 1.49 and 1.45 for AUC and MPL, respec- PL here: for the standard regimen, using the artesunate  killing tively, were observed (Figure 2; Supplementary Table 4). duration derived from Hendriksen’s PK parameters (Figure  3) Parameter analysis with PRCC (Supplementary Table  8) −5 we obtained a mean PRR of 5.18 (see Supplementary revealed that patients whose initial infections were in either Information for a nuanced discussion of PK parameters). very late or very early initial mean age-bins (Figure  5, lower panel) will have worse outcomes with the simplified regimen. Standard Regimen Treatment of Artemisinin-Sensitive Parasites This occurred because parasites in these stages are largely insen- We simulated treatment of drug-sensitive parasites with the sitive to artesunate at first treatment, and the simplified regimen standard regimen and identified the key drivers of pathology by lacks the second dose, 12 hours later, of the standard regimen calculating which parameters were most correlated with AUC PL that would effectively target these parasites that had matured and MPL (Figure 4; Supplementary Table 7). The most highly into more artemisinin sensitive age-bins. correlated parameter for both metrics was the initial parasite e h Th alf-life of the recovery rate r had a moderate correlation number: large positive PRCCs (between 0.88 and 0.98) were with outputs in the 12 to 24-hour and 24 to 48-hour periods, observed with associated P values ≤.001 at all time periods. The indicating that assumption of slower recovery made the simpli- half-life of the recovery rate r had PRCC of 0.46 for AUC and PL fied regimen perform relatively better (Supplementary Figure 5). 0.34 for MPL in the 24 to 48-hour time period (P values ≤.001), We are confident this parameter does not aeff ct the valid- but PRCC of <0.3 in earlier time periods. All other parameters ity of our results; for complete discussion see Supplementary had PRCC values of <0.3, indicating that outcome metrics were Information. No other parameters have notable correlation not highly correlated as per accepted statistical criteria [35]. All with sequestration-based pathology when comparing regimens. other model parameters had negligible correlation. The most We repeated this analysis to compare regimens (ie, WHO likely explanation is that such a large proportion of parasites standard vs simplified) when treating artemisinin-resistant par - are killed by artesunate that small differences in the number asites. Differences between regimens were extremely similar to killed are negligible compared to the initial parasite number those shown in Figure  5 and are displayed in Supplementary and pathological recovery rate. Figure 7 and Supplementary Table 9. Comparison of Simplified and Standard Regimen The Impact of Artemisinin Resistance on Treatment by the Standard We evaluated alternative treatment regimens on artemisi- Regimen nin-sensitive parasites. These results are presented as ratios of Unsurprisingly, ratios of AUC and MPL when comparing PL AUC and MPL. The simplified regimen had a slightly higher resistant and sensitive parasites are never less than 1 (Figure 2), PL Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1247 Value of outcome metric Ratio of outcome metric Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 123456 7 Duration of artesunate killing with 2.4 mg/kg [h] 2 34567 Duration of artesunate killing with 4 mg/kg [h] Figure 3. Distribution of artesunate killing duration. Data for 10 000 patients following treatment with a single dose of artesunate of either 2.4 mg/kg (A) or 4 mg/kg (B); note the duration includes that of the active metabolite dihydroartemisinin. This distribution was obtained using parameters from Hendriksen et al [33]. that is under no circumstance did patients have a better out- posttreatment largely coincides with parasites in age-bins come when parasites are resistant. Differences in median values insensitive to artesunate through resistance, rendering the ini- (Figure 2; Supplementary Table 4) were extremely small. tial dose nearly or completely ineffective. We carried out PRRC analysis (Supplementary Table  10) to SD of the initial mean age-bin had a positive correlation with investigate whether this small difference obscured the presence the ratio (indicating that resistant parasites had worse outcomes of a vulnerable subgroup of patients. This appeared to be the as SD increased). This occurred because higher SD “nudged” case: patients whose infections are clustered in the early age- parts of the age-bin distribution into (or out of) resistant bins at time of treatment had pathological outcomes that were age-bins (ie, the contiguous bin 45–48 and 1–5 where killing significantly worse in the presence of resistance (Figure 6). is absent). PRCC analysis showed no other parameter had a In these early age-bins, ratios for AUC and MPL are as PRCC value of >0.01, suggesting the initial mean age-bin (and, PL high as 5 in the 0 to 24-hour period (comparisons based on the to a lesser extent, its SD) are the sole determinants of whether upper quartile value). This occurs because artesunate presence a patient’s outcome will be worse in the presence of resistance. 1248 • JID 2019:219 (15 April) • Jones et al Counts Counts Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 1e+15 1e+13 1e+11 1e+09 10.0 10.5 11.0 11.5 12.0 10.0 10.5 11.0 11.5 12.0 Initial parasite numbers 10^ Figure  4. Analysis of the baseline scenario. The impact of underlying factors on the standard World Health Organization regimen used to treat patients with artemis- inin-sensitive parasites. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on values of area under the pathology curve (AUC ) and PL maximum pathological load (MPL) obtained from a population. B, Values of AUC and MPL are plotted against the most highly correlated parameter, that is initial parasite PL number, for 4 time periods posttreatment. DISCUSSION was highly flexible (discussed in Supplementary Information) and, of necessity, reflected the limitations in our understanding We established a PK/PD modeling methodology capable of of pathology, for example how rapidly pathology is resolved fol- investigating the treatment of severe malaria. Kremsner et  al lowing parasite death and whether pathology depends on max- [36] recognized the clinical necessity of this, and noted that imal sequestered load (measured as MPL) or on total exposure “for the first time, we [ie, Kremsner et al] are assessing artesu- (measured as AUC ). An interesting, highly important result is PL nate using similar pharmacokinetic and dynamic approaches”. that the key quantitative assumption made in the analysis, the Parasite clearance is likely to be a poor measure of regimen rate of resolution of pathology (measured as the half-life of r), had effectiveness (and, by extension, clinical outcome) in severe little effect on our conclusions when comparing alternative regi- malaria where pathology is due to sequestered parasites. The mens or the impact of resistance (Supplementary Information) effects of alternative regimens and the impact of drug resistance implying that the pathological model is a robust to assumptions can only be investigated by traditional clinical outcomes using made in this comparative investigation. Importantly, while cir- large-scale clinical trials, so pharmacological modeling of the culating parasite loads do not reflect the pathology of severe type proposed here is essential to help generate the evidence malaria they are currently the regular endpoint of choice in base for rational treatment design. Our pathological modeling Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1249 Value of outcome metric PRCC Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 110203040481 10 20 30 40 48 Mean age bin distribution [h] Figure 5. Evaluation of alternative drug treatment regimens. Comparison of the simplified versus World Health Organization standard regimen for treatment of artemis- inin-sensitive parasites; ratios of >1 indicate the simplified regimen produces worse outcome metrics. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on the ratios of area under the pathology curve (AUC ) and maximum pathological load (MPL). B, Ratios of AUC and MPL are plotted against the most PL PL highly correlated parameter (initial mean age-bin), for 4 time periods posttreatment. severe malaria trials, including those undertaken by Kremsner lower median AUC within the first 24 hours posttreatment PL et al [30, 37]; our model was able to reproduce the clinical out- (Figure  2; Supplementary Table 4). This difference was greater comes reported in [30, 33] (when appropriately parameterized), in the 24 to 48-hour period, but the majority of pathological and recover expected PRR , so we are confident it is reflective load occurred within the first 24 hours as artesunate rapidly of in vivo scenarios (Supplementary Information). kills parasites: AUC in the 24 to 48-hour period is, on aver- PL Kremsner and colleagues [30, 37] concluded that their sim- age, between 20% and 30% that of AUC in the 0 to 24-hour PL plified regimen was noninferior to the standard WHO regimen period (data not shown). The first 24 hours are critical for and possessed operational advantages due to less frequent drug patient survival [31], so outcome metrics at 24–48 hours may administration [30, 37]. This work was influential and ini- have little relevance in choosing between regimens. However, tiated a wider debate about the best drug regimen(s) to treat the simplified regimen performed much worse in the subgroup severe malaria [14, 36, 38] to which our study can contribute. of patients with very late or very early initial mean age-bins. Comparison of the 0 to 24-hour and 12 to 24-hour period was Based on these results, we are dubious about recommending used to compare the effects of the initial, larger dose of the sim- use of the simplified regimen but add an important rider to this. plified regimen against the additional dose at 12 hours with Kremsner et al never claimed this simplified regimen would be the standard regimen. The standard regimen produced slightly superior, but argued that any inferiority, if it exists, would be 1250 • JID 2019:219 (15 April) • Jones et al PRCC Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 0-12h 0-24h 12-24h 24-48h AUC MPL PL Parameters 1.0 Artesunate duration Half-life of r 1.5 Initial parasite number Initial mean age bin 0.0 Parasite multiplication rate –0.5 Standard deviation max –1.0 0-12 0-24 12-24 24-48 0-12 0-24 12-24 24-48 Time period [h] AUC MPL PL 110203040481 10 20 30 40 48 Mean age bin distribution [h] Figure 6. Analysis of the impact of artemisinin resistance. The effectiveness of the World Health Organization standard regimen used to treat resistant versus sensitive parasites; ratios of >1 indicate that resistant parasites have worse outcome metrics. A, Partial rank correlation coefficients (PRCC) using Spearman ρ of model parameters on the ratios of area under the pathology curve (AUC ) and maximum pathological load (MPL). B, Ratios of AUC and MPL are plotted against the most highly correlated PL PL parameter (mean age-bin), for 4 time periods posttreatment. within acceptable margins. We leave it to clinically qualified malaria (provided there was no resistance to partner drugs) personnel to judge whether 50% in some subgroups is within an but artemisinin resistance clearly poses a much larger threat acceptable margin of inferiority, especially given our inability to treatment of severe malaria than it does to uncomplicated to directly link our pathological outcomes with the likelihood malaria. Although differences between sensitive and resistant of mortality. parasites across the entire population are minor (Figure  2; We assessed the impact of artemisinin resistance on treat- Supplementary Table 4), there is an extremely vulnerable sub- ment of severe malaria, that is the extent to which resistance group of patients whose infections at the time of treatment are increased MPL and AUC . Resistance prevents drug killing in clustered in very late or very early age-bins (ie, where parasites PL age-bins 2–4 (these bins are otherwise hypersensitive) resulting are resistant in our model; Figure 6). in no killing for a contiguous 8-hour period in resistant par- Note that we specifically model relatively tightly synchro- asites (ie, age-bins 45 to 5). Our results show the initial mean nized parasite distributions (Supplementary Information); if age-bin and its SD are the only parameters that distinguish out- distributions were to become less tightly synchronized the vul- comes between sensitive and resistance parasites (Figure 6). We nerability of patients with early initial mean age-bins decreases argued previously [39] that artemisinin resistance would have and both the difference between regimens and the impact of a negligible impact on eventual cure rates in uncomplicated resistance reduces. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1251 PRCC Ratio of outcome metric Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 We present a highly adaptable methodology for PK/PD mod- 6. Dondorp AM, Kager PA, Vreeken J, White NJ. Abnormal eling of treatment of severe malaria that was able to recover key blood flow and red blood cell deformability in severe clinical observations (based on circulating parasite numbers), malaria. Parasitol Today 2000; 16:228–32. and, with novel metrics, used to investigate the pathology of 7. Clark IA, Alleva LM. Is human malarial coma caused, or severe malaria. Our model showed that while on a popula- merely deepened, by sequestration? Trends Parasitol 2009; tion level a simplified artesunate regimen is noninferior to the 25:314–8. standard WHO regimen, outcomes in a subgroup of patients 8. Medana IM, Turner GD. Plasmodium falciparum and the with infections grouped in late or early initial mean age-bins blood-brain barrier–contacts and consequences. J Infect are notably worse with the simplified regimen. The emergence Dis 2007; 195:921–3. of artemisinin resistance in early ring stages poses a significant 9. Smith JD, Rowe JA, Higgins MK, Lavstsen T. Malaria’s threat to this same group of patients. Neither of these results are deadly grip: cytoadhesion of Plasmodium falciparum-in- particularly obvious from summary statistics of the population fected erythrocytes. Cell Microbiol 2013; 15:1976–83. and so subgroup analysis is particularly important in devising 10. Hughes KR, Biagini GA, Craig AG. Continued cytoadherence treatment strategies for severe malaria. of Plasmodium falciparum infected red blood cells aer a ft nti- malarial treatment. Mol Biochem Parasitol 2010; 169:71–8. Supplementary Data 11. Udomsangpetch R, Pipitaporn B, Krishna S, et  al. Supplementary materials are available at e Th Journal of Infectious Antimalarial drugs reduce cytoadherence and rosetting Diseases online. Consisting of data provided by the authors to Plasmodium falciparum. J Infect Dis 1996; 173:691–8. benefit the reader, the posted materials are not copyedited and 12. van Hensbroek MB, Onyiorah E, Jaffar S, et  al. A trial of are the sole responsibility of the authors, so questions or com- artemether or quinine in children with cerebral malaria. N ments should be addressed to the corresponding author. Engl J Med 1996; 335:69–75. 13. Esu E, Effa EE, Opie ON, Uwaoma A, Meremikwu MM. Notes Artemether for severe malaria. Cochrane Database Syst Rev Acknowledgment. We thank 2 anonymous reviewers for 2014; (9):CD010678. their helpful comments and critique. 14. Dondorp AM, Maude RJ, Hendriksen IC, Day NP, White Financial support. This work was supported by the UK NJ. Artesunate dosing in severe falciparum malaria. J Infect Medical Research Council (grant numbers G1100522 and MR/ Dis 2012; 206:618–9; author reply 622. L022508/1); the Bill and Melinda Gates Foundation (grant 15. Hastings IM, Kay K, Hodel EM. How robust are malaria par- number1032350); and the Malaria Modeling Consortium asite clearance rates as indicators of drug effectiveness and (grant number UWSC9757). resistance? Antimicrob Agents Chemother 2015; 59:6428–36. Potential coni fl cts of interest. All authors: No reported con- 16. Ferreira PE, Culleton R, Gil JP, Meshnick SR. Artemisinin flicts of interest. All authors have submitted the ICMJE Form resistance in Plasmodium falciparum: what is it really? for Disclosure of Potential Conflicts of Interest. Conflicts that Trends Parasitol 2013; 29:318–20. the editors consider relevant to the content of the manuscript 17. Krishna S, Kremsner PG. Antidogmatic approaches to arte- have been disclosed. misinin resistance: reappraisal as treatment failure with arte- misinin combination therapy. Trends Parasitol 2013; 29:313–7. References 18. Hastings IM, Hodel EM, Kay K. Quantifying the pharma- cology of antimalarial drug combination therapy. Sci Rep 1. World Health Organization (WHO). World malaria report 2016; 6:32762. 2016. Geneva, Switzerland: WHO, 2016. 19. Hodel EM, Kay K, Hastings IM. Incorporating stage-spe- 2. Olliaro P. Editorial commentary: mortality associated with cific drug action into pharmacological modeling of antima- severe Plasmodium falciparum malaria increases with age. larial drug treatment. Antimicrob Agents Chemother 2016; Clin Infect Dis 2008; 47:158–60. 60:2747–56. 3. Cunnington AJ, Riley EM, Walther M. Stuck in a rut? 20. Kay K, Hastings IM. Improving pharmacokinetic-phar- Reconsidering the role of parasite sequestration in severe macodynamic modeling to investigate anti-infective che- malaria syndromes. Trends Parasitol 2013; 29:585–92. motherapy with application to the current generation of 4. Mackintosh CL, Beeson JG, Marsh K. Clinical features antimalarial drugs. PLoS Comput Biol 2013; 9:e1003151. and pathogenesis of severe malaria. Trends Parasitol 2004; 21. Kay K, Hodel EM, Hastings IM. Altering antimalarial drug 20:597–603. regimens may dramatically enhance and restore drug effec- 5. Taylor T, Olola C, Valim C, et al. Standardized data collec- tiveness. Antimicrob Agents Chemother 2015; 59:6419–27. tion for multi-center clinical studies of severe malaria in 22. Hodel EM, Kay K, Hastings IM. Incorporating stage-spe- African children: establishing the SMAC network. Trans R cific drug action into pharmacological modeling of Soc Trop Med Hyg 2006; 100:615–22. 1252 • JID 2019:219 (15 April) • Jones et al Downloaded from https://academic.oup.com/jid/article/219/8/1243/5231587 by DeepDyve user on 13 July 2022 antimalarial drug treatment. Antimicrob Agents Chemother 31. Kendjo E, Agbenyega T, Bojang K, et al. Mortality patterns 2016; 60:2747–56. and site heterogeneity of severe malaria in African children. 23. R Core Team. R: A language and environment for statisti- PLoS One 2013; 8:e58686. cal computing. Vienna, Austria: R Foundation for Statistical 32. Maitland K. Management of severe paediatric malaria in Computing, 2014. resource-limited settings. BMC Med 2015; 13:42. 24. Chua CL, Brown G, Hamilton JA, Rogerson S, Boeuf P. 33. Hendriksen IC, Mtove G, Kent A, et al. Population pharma- Monocytes and macrophages in malaria: protection or cokinetics of intramuscular artesunate in African children pathology? Trends Parasitol 2013; 29:26–34. with severe malaria: implications for a practical dosing reg- 25. White NJ. The parasite clearance curve. Malar J 2011; imen. Clin Pharmacol Ther 2013; 93:443–50. 10:278. 34. White NJ. Assessment of the pharmacodynamic prop- 26. White NJ. Malaria parasite clearance. Malar J 2017; 16:88. erties of antimalarial drugs in vivo. Antimicrob Agents 27. Saralamba S, Pan-Ngum W, Maude RJ, et al. Intrahost mod- Chemother 1997; 41:1413–22. eling of artemisinin resistance in Plasmodium falciparum. 35. Hinkle DE, Wiersma W, Jurs SG. Applied statistics for the Proc Natl Acad Sci U S A 2011; 108:397–402. behavioral sciences. 5th ed. Boston: Houghton Mifflin, 2003. 28. Gordi T, Xie R, Jusko WJ. Semi-mechanistic pharmaco- 36. Kremsner PG, Taylor T, Issifou S, et al. A simplified intrave- kinetic/pharmacodynamic modelling of the antimalarial nous artesunate regimen for severe malaria–reply. J Infect effect of artemisinin. Br J Clin Pharmacol 2005; 60:594–604. Dis 2012; 206:622. 29. Hietala SF, Martensson A, Ngasala B, et  al. Population 37. Kremsner PG, Taylor T, Issifou S, et al. A simplified intra- pharmacokinetics and pharmacodynamics of artemether venous artesunate regimen for severe malaria. J Infect Dis and lumefantrine during combination treatment in chil- 2012; 205:312–9. dren with uncomplicated falciparum malaria in Tanzania. 38. Woodrow CJ, Taylor WR. Reappraisal of the efficacy of a Antimicrob Agents Chemother 2010; 54:4780–8. simplified artesunate regimen in falciparum malaria. J 30. Kremsner PG, Adegnika AA, Hounkpatin AB, et  al. Infect Dis 2012; 206:619–21. Intramuscular artesunate for severe malaria in African chil- 39. Hastings IM, Hodel EM, Kay K. Quantifying the pharma- dren: a multicenter randomized controlled trial. PLoS Med cology of antimalarial drug combination therapy. Sci Rep 2016; 13:e1001938. 2016; 6:32762. Modeling Severe Malaria Treatment • JID 2019:219 (15 April) • 1253

Journal

The Journal of Infectious DiseasesOxford University Press

Published: Apr 8, 2019

Keywords: pathology; artemisinine; artesunate; malaria, falciparum, severe and complicated; parasites; world health organization; infections

There are no references for this article.