key: cord-0254407-leo7j7r3
authors: Burgert, L.; Reiker, T.; Golumbeanu, M.; Moehrle, J. J.; Penny, M. A.
title: Model-informed target product profiles of long acting-injectables for use as seasonal malaria prevention
date: 2021-07-05
journal: nan
DOI: 10.1101/2021.07.05.21250483
sha: 835c1fd3a3bf0c03e00f9bd4fbd6f42c089ea987
doc_id: 254407
cord_uid: leo7j7r3

Seasonal malaria chemoprevention (SMC) has proven highly efficacious in reducing malaria incidence. However, the continued success of SMC is threatened by the spread of resistance against one of its main preventive ingredients, Sulfadoxine-Pyrimethamine(SP), operational challenges in delivery, and incomplete adherence to the regimens. Via a simulation study with an individual-based model of malaria dynamics, we provide quantitative evidence to assess long-acting injectables (LAIs) as potential alternatives to SMC. We explored the predicted impact of a range of novel preventive LAIs as a seasonal prevention tool in children aged three months to five years old during late-stage clinical trials and at implementation. LAIs were co-administered with a blood-stage clearing drug once at the beginning of the transmission season. We found the establishment of non-inferiority of LAIs to standard 3 or 4 rounds of SMC with SP-amodiaquine was challenging in clinical trial stages due to high intervention deployment coverage. However, our analysis of implementation settings where the achievable SMC coverage was much lower, LAIs with fewer visits per season are potential suitable replacements to SMC. Suitability as a replacement with higher impact is possible if the duration of protection of LAIs covered the duration of the transmission season. Furthermore, optimizing LAIs coverage and protective efficacy half-life via simulation analysis in settings with an SMC coverage of 60% revealed important trade-offs between protective efficacy decay and deployment coverage. Our analysis additionally highlights that for seasonal deployment for LAIs, it will be necessary to investigate the protective efficacy decay as early as possible during clinical development to ensure a well-informed candidate selection process.

Seasonal malaria chemoprevention (SMC) has proven highly efficacious in reducing malaria incidence. 12 However, the continued success of SMC is threatened by the spread of resistance against one of its main 13 preventive ingredients, Sulfadoxine-Pyrimethamine(SP), operational challenges in delivery, and 14 incomplete adherence to the regimens. Via a simulation study with an individual-based model of malaria 15 dynamics, we provide quantitative evidence to assess long-acting injectables (LAIs) as potential alternatives 16 to SMC. We explored the predicted impact of a range of novel preventive LAIs as a seasonal prevention 17 tool in children aged three months to five years old during late-stage clinical trials and at implementation. 18

LAIs were co-administered with a blood-stage clearing drug once at the beginning of the transmission 19 season. We found the establishment of non-inferiority of LAIs to standard 3 or 4 rounds of SMC with SP-20 amodiaquine was challenging in clinical trial stages due to high intervention deployment coverage. 21

However, our analysis of implementation settings where the achievable SMC coverage was much lower, 22

LAIs with fewer visits per season are potential suitable replacements to SMC. Suitability as a replacement 23

with higher impact is possible if the duration of protection of LAIs covered the duration of the transmission 24 season. Furthermore, optimizing LAIs coverage and protective efficacy half-life via simulation analysis in 25 settings with an SMC coverage of 60% revealed important trade-offs between protective efficacy decay 26 and deployment coverage. Our analysis additionally highlights that for seasonal deployment for LAIs, it 27

will be necessary to investigate the protective efficacy decay as early as possible during clinical 28 development to ensure a well-informed candidate selection process. 29 30 3 2. Background 31 Children carry the majority of the global malaria burden, with an estimated 67% of all malaria-related 32 deaths (272 000 in 2019) occurring in those under 5 years of age 1 . In addition to effective and timely 33 treatment, prevention through vector control or drug-based prophylaxis has proven to be an effective 34 approach, reducing incidence and mortality 2 . Especially in seasonal malaria transmission settings, where 35 malaria transmission is linked to the rainy months, well-timed preventive malaria interventions that protect 36 from infection during the transmission months can ease a substantial amount of malaria burden 1 . The WHO 37 has recommended seasonal malaria chemoprevention (SMC) with monthly Sulfadoxine-38

Pyrimethamine+Amodiaquine (referred to as SMC-SP+AQ) for children aged between 3 months and 5 39 years during the malaria transmission season since 2012 3 . SP+AQ provides a two-stage effect: while AQ 40

clears existing blood-stage infections, the long clearance half-life of SP prevents new infections. The impact 41

of SMC in seasonal settings has been widely demonstrated, achieving a protective efficacy of roughly 80% 42

against clinical episodes in a trial in Burkina Faso 4 , a reduction in incidence by 60% in routine 43 implementation in Senegal (80% deployment coverage of all eligible children) 5 , and a reduction in the 44 proportion of positive tests by 44% in routine implementation in Mali 6 . 45

Despite its potential, poor adherence and the spread of drug resistance limit the effectiveness of SMC. 46 Additionally, the monthly delivery of SMC-SP+AQ (one day of SP and three days of AQ) throughout the 47 transmission season is relatively expensive, due to human resources and especially due to operational 48 constraints during the rainy season 7 . Consequently, in 2019, only 62% of children living in SMC-eligible 49 areas in the Sahel subregion received SMC 1 . Throughout the transmission season, coverage subsequently 50 decreased by 6% in Guinea 8 and 20% in Mali 9 . Investigation of the adherence to the 3x AQ regimen within 51 one treatment course in a clinical trial in Niger showed that only 20% of children received the full regimen 52 10 . Additionally, the spread of resistance markers against SP was reported with increasing SMC deployment 53 11, 12 , impacting the eligibility of regions for SMC 13 as well as the protective efficacy after implementation 54 14 . 55

The need for alternative prevention tools that simplify deployment and possess a reduced risk of resistance 56 is pressing. In the absence of an effective vaccine, long-acting injectables (LAIs) with an anti-infective 57 effect could provide an alternative seasonal prevention tool by simplifying the deployment and reducing 58 the risk of resistance through decreased selection pressure for SP resistance 15 . Current candidate LAIs 59 include small molecule drugs 16, 17 or monoclonal antibodies (mAbs) 18,19 that target the sporozoite stage or 60 liver stage of the malaria parasite, thereby serving as anti-infectives. The successful development of a LAI 61 entails the definition of appropriate product profiles and use cases which are specified in Target Product  62 Profiles (TPPs). Precisely, these specifications include LAI efficacy and safety prerequisites as well as 63 delivery modalities 15 . TPPs are living documents and therefore continuously updated as new evidence for 64 product requirements becomes available. 65

To justify the implementation of LAIs under the use case of seasonal malaria prevention, non-inferiority to 66 the existing intervention of SMC-SP+AQ must be met 4 . For new tools with new modes of action and/or 67 deployment modalities, proving non-inferiority to the standard of care is a crucial step and is usually 68 established in non-inferiority clinical trials conducted at late stages of tool development. Currently, it is not 69 yet known what clinical studies will be required for LAIs. In absence of efficacy data on LAIs, it is at the 70 current stage impossible to obtain insights about the circumstances in which LAIs have the potential to be 71 non-inferior. In silico modelling and simulation approaches of malaria transmission and control, allow the 72 quantification of the impact of varying tool specifications in relation to varying operational and setting 73 constraints which would not be feasible in real life clinical trials. They thus allow for the exploration of the 74 potential to meet non-inferiority criteria. In the absence of non-inferiority evidence at the early stages of 75 development, modelling and simulation approaches therefore provide a quantitative basis to further inform 76 decision making. They guide tool development from the early stages by predicting potential public health 77 impact (Fig. 1a ) 20 and understanding non-inferiority criteria prior to clinical trial planning. 78

Here, we investigate the potential public-health impact for various use cases of LAIs by conducting an in 79 silico simulation analysis examining product properties and operational modalities supporting LAI 80 implementation as a seasonal malaria prevention tool. Accordingly, in the simulated scenarios, LAIs were 81 delivered to children under five once at the beginning of the transmission season with an antimalarial in 82 settings currently approved for SMC deployment. Their protective effect was then compared to SMC-83 SP+AQ administered three or four times per season in monthly intervals. By combining disease modelling 84

and simulation experiments with machine learning approaches, we efficiently explored the large space of 85

possible parameter values describing intervention and transmission setting characteristics and analyzed 86

trade-offs between tool characteristics and operational constraints in a variety of transmission settings 20 . 87 We conducted our malaria transmission simulations using OpenMalaria, an established individual-based 88 stochastic simulation platform of malaria transmission and control 21 22 . Based on this approach, we defined 89 a quantitative framework to assess the influence of tool properties and operational constraints on the impact 90 of SMC and LAIs. Using this framework, we investigated two malaria transmission settings based on the 91 malaria transmission profile in Senegal and Mali and assessed public health impact for a plethora of 92 different tool properties, deployment coverages and over multiple transmission intensities. Our analysis 93 was carried out along the clinical development pathway from late clinical trials through to implementation 94 of future LAIs as an SMC replacement. By understanding the main drivers of impact to reach a pre-defined 95 health goal in implementation stages and under operational constraints, we provide an assessment of 96 endpoints in clinical trials of newly developed LAIs and identify efficacy requirements for further 97 development. 98

The impact of novel anti-infective LAIs depends on the tool properties defining their efficacy profile, as 100 well as on the operational constraints and the respective underlying malaria epidemiology, that influence 101 tool suitability for implementation in a given setting (Fig. 1a) . We investigated the influence of tool 102

properties over a large range of specified protective efficacies, as well as operational considerations 103

(coverage of children) in several implementation settings varying in seasonality and access to healthcare. 104

The drivers of predicted impact for preventive LAIs were analysed along their clinical development from 105 clinical trials to implementation (Fig. 1b ) and compared to current standard of care (SMC-SP+AQ). 106

Accordingly, we defined two analysis stages: in the clinical trial stage, we investigated the maximum 107 incidence reduction and the ability to establish non-inferiority to SMC-SP+AQ over one malaria season 108 ( Fig. 1 b, panel 1 ). In the implementation stage, we replaced SMC-SP+AQ with LAIs after five years of 109 implementation at varying coverages, and we inferred the minimal LAI coverage at which LAIs are 110 equivalent (non-inferiority) in public-health impact (incidence reduction) to a continued implementation of 111 SMC-SP+AQ (Fig. 1b, panel 2) . 112 modality, tool specifications, and settings in which a concrete health target was analysed (in our case, 115 incidence reduction). Second, a set of disease scenarios were simulated randomly over the entire parameter 116 space to evaluate the health outcomes. The resulting database of simulations was used to train a Gaussian 117 process emulator (GP), that predicts the health outcome given a set of input parameters. Third, the emulator 118 was employed to perform sensitivity analysis and optimisation of tool properties with respect to health 119

outcomes. This analysis allowed us to define the optimal product characteristics of a LAI that maximises 120 the chance of achieving a desired health goal. 121

We explored the dynamics of a preventive LAI against malaria using OpenMalaria, a stochastic, individual-123 based simulator of malaria infection in humans linked to a deterministic model of the mosquito feeding 124 cycle 23,24 . OpenMalaria accounts for heterogeneity within a human population on multiple levels including 125 host exposure, susceptibility and immune response 21,25,26 . The model allows the investigation of 126

interventions against malaria at multiple points in the malaria life cycle (e.g. vaccines 22 , insecticide treated 127 bed-nets, and reactive case detection 27 ) while monitoring a variety of health outcomes (e.g. prevalence, 128

incidence, mortality) 28 . 129

Using OpenMalaria, we simulated a range of transmission settings (Fig. S1 ) and assumptions for the 131 implementation of SMC and LAIs as a seasonal infection prevention intervention replacing existing 132 prevention with SMC-SP+AQ. These assumptions are with regards to the properties of the setting 133 (seasonality and intensity of transmission), health system (access to care and treatment of clinical cases), 134 new and replacement intervention (Table 1) . 135

The intervention age-group consisted of children between 3 months and 5 years of age (accounting for ca. 136

16% of the total population). The intervention age-group was chosen according to WHO recommendations 137 3 , although some countries have implemented SMC for children up to 10 years old 29 . We assumed a total 138 population of 10,000 individuals to capture transmission settings within health facility catchment areas with 139 an age-structure that represents a realistic age-distribution for African malaria-endemic settings 30 . Access 140 to treatment, defined as the percentage of the whole population who seek care for a symptomatic malaria 141 episode, was chosen to represent settings with low or high health systems strength. The probability of 142 symptomatic cases to receive effective treatment within two weeks from the onset of symptoms (E14) was 143 set to 10% in low access to health-care settings and 50% in high access to health-care settings 31 . The 144 malaria seasonality profile, mosquito species and timing of interventions were parameterised to reflect those 145 of Mali or Senegal, two countries in the Sahel region where SMC is implemented and clinical trials for 146 malaria interventions are conducted frequently. In Mali, the malaria season is longer, starting in August and 147 lasting until November (long season), and SMC is generally four monthly doses. In contrast, the malaria 148 season in Senegal is only three months long, with a sharper profile (short season) and SMC is three doses 149 one month apart (Fig. 1a, Table 1 , Figure S2 ). Malaria prevalence in Senegal is generally low, with the 150 highest P. falciparum prevalence in 2-10 years old (PfPR2-10y) in the southern regions being around 8%. 151

However, the PfPR2-10y in Mali is around 80% in the south of Mali but very low in the North 32 . 152

To develop a broader understanding of the influence of transmission intensity on LAI impact, we simulated 153 a range of initial incidence settings capturing the transmission heterogeneity of these two malaria-endemic 154 countries ( Figure S1 , Table S1 ) . The simulated transmission intensity of each setting was defined by the  155  entomological inoculation rate (EIR) modelled as the average annual number of infectious bites received  156  by an individual, and the corresponding simulated PfPR2-10y (Table S1 , Fig. S1 ). The protective efficacy of 157 SMC-SP+AQ was implemented as being fully effective (no prevalent resistance against SP) or reduced 158 duration of protection (prevalent resistance against SP). 159

Over all simulation experiments, the input parameter space for the protective efficacy and its decay, and 160 the intervention coverage are as defined in Table 1 The clinical trial stage (Fig. 1b, panel 1 ) was simulated in the high case management setting to account for 175 awareness of malaria symptoms. Initial deployment coverage levels for both, SMC and LAI, were set to 176 100%. Follow up visits, in the form of active case detection, were implemented two weeks after every 177 administration round of SMC-SP+AQ in both trial arm. In addition to fully effective SMC-SP+AQ, we 178 investigated a reduced length of protection by prevalent resistance against SP in the clinical trial stage. 179

In the implementation stages (Fig. 1b, panel 2) , we analysed the impact of switching seasonal malaria 180 prevention strategies from SMC-SP+AQ after five years of implementation to LAIs. After five years of 181 LAI implementation, the cumulative clinical case incidence was compared to a control setting (no 182 switching). LAIs and SMC-SP+AQ were implemented with varying coverages between 40-100 %. 183

Intervention cohorts, defined by the coverage in the intervention age group, were specified at the beginning 184 of a transmission season for the whole season. An exemplary illustration of LAI and SMC-SP+AQ 185 implementation is provided in Fig. S4 . 186

As the decay of efficacy of LAIs is not yet known, we explored a range of possible scenarios. In both the 188 clinical and implementation analysis, the prevention of infection by LAI was modelled as pre-erythrocytic 189 protection from infection E(t) over time defined by the initial protective efficacy E0 [%], protective efficacy 190

half-life of decay L and shape parameter k (Fig. S3) 

Eq Information. Scenarios assuming SP resistance, the effect of resistance against the SMC component SP was 209

implemented by decreasing the SMC half-life of protection from 32 days to 20 days. The lower protection 210 half-life due to resistance can be modeled as an increase in the drug concentration inhibiting the parasite 211 growth by 50% (IC50). Because SP has a long clearance half-life 40 , we assume no impact of resistance on 212 the initial efficacy. 213

Health target: endpoints to assess impact of LAI 214

The health targets analysed in this study were based on incidence reduction by LAIs, clinical cases averted 215

by LAIs compared to SMC, and non-inferiority with regard to clinical burden in the modelled clinical trials. 216

Clinical cases per person in the intervention age group (cpp0.25-5y, int) over the clinical trial length (Fig. 1a ) 217 (cpp0.25-5y, int) or in the last implementation year (cpppy0.25-5y, int) ( Fig. 1b) were calculated over the whole 218 population at risk in the intervention age-group (Nint,0.25-5y). 219

The incidence reduction percentage incred was calculated via the cumulative incidence in the year before 221 trial implementation cppcont and the cumulative incidence during the clinical trial cppint 222 to be non-inferior to SMC-SP+AQ if the upper limit of the derived 95% confidence interval of the hazard 226 ratio between SMC-SP+AQ and LAI lies below the upper limit for non-inferiority. This limit is informed 227 by the survival estimate of SMC-SP+AQ and the desired margin of non-inferiority. We assumed a margin 228 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. (Table S2) . 247

We identified the main drivers of intervention impact via a global sensitivity analysis, performed using 249 decomposition of variance via Sobol analysis on the emulator output predictions. We conducted an EIR-250 stratified sensitivity analysis to assess the potential change in drivers of impact over the whole transmission 251 range. Within the pre-defined parameter space, the total effect indices quantify the interactions between 252

individual parameter contributions to the emulator output variance 43 . The total effect indices were 253 normalized to obtain the relative importance of each parameter through division by their sum. The total 254 effect indices were estimated with a Monte Carlo sampling approach using the function soboljansen in the 255 R-packages sensitivity (version 1.16.2) with 500 000 sampled points and 1000 bootstrap replicates. 256

In the clinical trial stage, we discretized the space of the initial protective efficacy and protective efficacy 258

half-life of LAIs within the range of plausible values (Table 1) , yielding a two-dimensional grid of 259 parameter value combinations. At each grid point, we used the GP emulators to estimate the upper limit of 260 the confidence interval of the hazard ratio between the survival estimates in the SMC and LAI arm and the 261 non-inferiority margin and check if non-inferiority could be established for the given combination of 262 parameter values (cf. Non-inferiority Analysis in Supplementary Material). The contours of the resulting 263 non-inferiority surface yielded the thresholds of the minimal initial protective efficacy and protective 264 efficacy half-life needed to establish non-inferiority across different transmission settings (Fig. 3) . 265

In the implementation stage, we additionally determined the minimum required LAI coverage and half-life 266 at which non-inferiority to SMC-SP+AQ could be established. At each grid point, we conducted a 267 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint constrained optimization, translating the non-inferiority condition into an inequality constraint by requiring 268 the difference between the upper limit of the confidence interval and the non-inferiority margin to be 269 positive. The optimization was conducted using an augmented Lagrange method (gosolnp, R-package 270

Rsolnp, Version 1.16) with a minimum of 3 starting values and 200 simulations. To determine the benefits 271 of reallocation of resources from reduced visits within a season towards increasing deployment coverage, 272

we compared the number of cases per person per year in a simulation with SMC implemented at a given 273

coverage to the number of cases per person per year with LAI at a coverage that was 20% higher than the 274 optimal coverage. 275 276 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint

Decay shape and duration of protective efficacy influence LAI impact 278

We assessed the obtained malaria incidence reduction in the simulated scenarios of the clinical trial stage 279 at high deployment coverage (100%) and found that the decay shape of LAI protective efficacy and 280 protection half-life play an important role in achieving a targeted incidence reduction (Fig. 2) in each 281 simulated clinical trial scenario. Additionally, we identified the parameter space in which a certain 282 incidence reduction cannot be achieved following LAI deployment under different setting assumptions 283 (below each curve in Fig. 2 ). 284

In settings with a more extended transmission season (Fig. 2 b, d and f), a longer LAI half-life is required 285 to reach the same predicted impact compared to the shorter season settings (Fig. 2 a, c, and e). A steeper 286 initial decrease of initial protective efficacy ( Fig. 1a and Fig. S3 ) led to lower estimated incidence reduction, 287

with bi-phasic LAIs exhibiting the lowest predicted impact ( ( Fig. 2 b , d, and f) was predicted to increase the half-life requirements to reach a desired incidence reduction 295 for all LAIs. In these longer transmission season settings, a predicted incidence reduction of over 80% was 296 not possible for exponential and bi-phasic LAIs. 297 Fig. 2 shows the exemplary extraction of minimum essential efficacy properties for a LAI with a given half-298

life. For example, if the half-life of protective efficacy of a LAI was assumed to be 150 days, we predicted 299 that an initial protective efficacy of 88%, 96%, and 76% was required for exponential, bi-phasic and 300

sigmoidal LAIs, respectively (Fig. 2 a, c and e) to reach a clinical incidence reduction of 60% (short malaria  301 transmission season, initial cases per person per year0.25-5y of 2.8). 302

In the SMC-SP+AQ arm of the simulated clinical trial stages, we found a predicted mean achievable 304

incidence reduction of approximatively 90 % in Mali and Senegal-like seasonal settings (Table S 3) . Our 305 non-inferiority analysis (Fig. 1b, panel 3 ) demonstrated that the predicted establishment of non-inferiority 306

of sigmoidal LAIs to SMC-SP+AQ under the assumption of 100% initial deployment coverage could only 307 be met with LAI efficacy over 90% in both seasonal settings and half-life over 62 days in Senegal-like 308 (short season) and 88 days in Mali-like (long season) seasonal transmission patterns (Fig. 3) . In agreement 309 with the analysis of attainable incidence reduction (Fig. 2) , the predicted establishment of non-inferiority 310 was more feasible in settings with a shorter transmission season and lower initial malaria incidence ( Fig. 3  311 a). For settings with a lower initial incidence (between 0.5 and 1 initial cases per person per year0.25-5y), the 312 parameter space where non-inferiority could be established varied more than in higher initial incidence 313

settings. If resistance against SP was prevalent, which we modelled as a shorter duration of protection 314 through a decrease in protective efficacy half-life to 20 days (from 32 days), sigmoidal LAIs were predicted 315 to be non-inferior in a wider range of tool property combinations (Fig. 3 , c and d). Nevertheless, non-316 inferiority could not be established in any setting for any parameterization of exponential and bi-phasic 317 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021.

LAIs under clinical trial coverage assumptions. We conclude that the efficacy decay profiles of LAIs play 318 an important role for reaching the defined incidence reduction goals and establishing non-inferiority to 319 SMC-SP+AQ. 320

Moving from clinical field trials towards implementation stages, where LAIs are administered as a 322

replacement for SMC-SP+AQ ( Fig. 1 b, panel 2) , we analyzed the influence of underlying LAI efficacy 323

properties and deployment coverage on resulting intervention impact and non-inferiority to SMC-SP+AQ. 324

Our sensitivity analysis via decomposition of variance ( Fig. 4) attributable variance, accounting for 38 % to 52% of variance while the initial protective efficacy 333

contributed between 7% to 19%, depending on transmission intensity. However, for half-lives greater than 334 90 days (Fig. 4 b) the importance of deployment coverage increased from 54% to 85% and importance of 335 the initial protective efficacy from 14% to 38 % (in these results we assumed LAIs of less than 70% initial 336 protective efficacy are unlikely to be developed). In contrast, the relative importance of the protective 337 efficacy half-life decreased to around 2% to 12%. In settings with a longer malaria transmission season 338 ( Fig. S6 a, d, g), a sharper initial decrease in clinical incidence was predicted for a larger range of protective 339 efficacy half-life than for shorter transmission seasons ( Fig. S5 a, d, g). Overall, this demonstrates the 340 potential to explore how the impact determinants and their importance change based on efficacy duration 341 cut-offs compared to length of transmission season or alternative deployment. 342

These results illustrate the importance of setting-specific trade-offs between enhancing tool properties or 344 improving implementation coverage (Table 2) . For example, increasing the half-life of a sigmoidal LAI 345

with an initial efficacy of 90% from 49 days to 63 days reduced the predicted required LAI coverage to 346 establish non-inferiority to SMC-SP+AQ in implementation (60 % coverage) by 20 % (from 100 % to 80 347 %) in a setting with an initial clinical incidence of 1.4 cases per person per year0.25-5y. Furthermore, in 348 settings with relatively high levels of initial clinical malaria incidence and corresponding high transmission 349 intensity, namely cases per person per year0.25-5y>2.4 and EIR>150, a change in dynamics to establish non-350 inferiority was observed. In these settings, we predicted LAIs will likely fail to sufficiently protect the 351 targeted population from clinical malaria even at very high deployment coverage (Fig. S4) . Therefore, we 352

were unable to assess the required half-life of protective efficacy of LAI for high transmission settings. 353

For SMC-SP+AQ, we estimated that an additional 13 % to 29 % in incidence reduction could be achieved 354 by increasing the coverage from 62% to 100%, dependent on initial clinical incidence before 355 implementation (Fig. S7) . However, achieving high levels of SMC coverage at implementation is 356 challenging 5,6 , and increasing levels of coverage are associated with increasing costs. 357 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint

As information on costs of LAI (costs of goods and supply chain) are not available as of now, we were 358 unable to include detailed economic analysis in the assessment of LAIs in implementation stages. The main 359 cost drivers of SMC-SP+AQ are deployment costs (remuneration of health care workers) and cost of goods 360 44,45 , with deployment costs increasing non-linearly with higher coverage. Therefore, we determined the 361 minimal LAI coverage at which non-inferiority to SMC-SP+AQ (assuming initial SMC deployment 362 coverage of 60%) was established, stratified by initial clinical incidence before implementation (Fig 5 and  363 Fig S8 and S9 ). We found that the parameter space where non-inferiority could be established shrank with 364

increasing baseline malaria incidence ( Fig. S8 and S9 ). With regard to seasonality, LAIs were more likely 365

to be non-inferior in shorter malaria transmission settings in the implementation stage ( Fig. S7 and Fig. S8 ). 366

In settings with a high initial clinical incidence (EIR=150 , Table S1 ), LAI coverage could not be optimized 367

to be non-inferior to SMC-SP+AQ at 60% coverage because LAIs were unable to prevent malaria cases 368 even at very high coverage levels (Fig. S4d) . Overall, the optimisation of LAI coverage in settings with a 369

limited SMC-SP+AQ coverage illustrates the potential of LAI implementation in non-optimal coverage 370 settings. 371

While non-inferiority could only be established in a small part of the parameter space of sigmoidal LAIs in 372 long seasonal transmission settings (Fig. 3b) , it is possible to increase the potential area of applicability of 373

sigmoidal LAIs by optimising their deployment coverage (Fig. 5a , initial cases per person per year0.25-374 5y=1.4). Deploying a sigmoidal LAI at 46% coverage with a half-life of 150 days and initial efficacy of 375

100% is sufficient to establish non-inferiority over SMC-SP+AQ at 60% coverage (Fig. 5 b) . In contrast, 376

sigmoidal LAIs with a half-life of 70 days and initial efficacy of 70 % require a deployment coverage of 377 95% in order to be non-inferior (Fig. 5 e) . For sigmoidal LAIs we found that increasing the deployment 378 coverage over the estimated minimum required coverage to establish non-inferiority results in potential 379 gains in terms of clinical incidence reduction compared to SMC-SP+AQ ( Fig. 5 b to e). 380

Additionally, we found that even though non-inferiority of exponential LAIs to SMC-SP+AQ could not be 381 established in clinical trial stages, coverage optimisation in implementation stages reveals their 382 applicability. Deploying an exponential LAI at 52% coverage with a half-life of 150 days and initial efficacy 383 of 100% was sufficient to establish non-inferiority over SMC-SP+AQ at 60% coverage (Fig. S10b ) and a 384

half-life of 100 days and initial protective efficacy of 90% requires a deployment of 78% to establish non-385 inferiority (Fig. S10c) . Exponential LAIs with a half-life of 70 days and initial efficacy of 70 % were always 386 inferior to SMC-SP+AQ at a coverage of 60% (Fig. S10e) . 387

The effective prevention of clinical malaria in children is crucial to prevent malaria mortality and reduce 389 the overall global malaria burden 1 . Through modelling and simulation, we explored a broad range of LAI 390 characteristics in multiple settings for clinical testing and deployment. This allowed us to understand the 391 likely influence of LAI efficacy properties and operational factors on clinical incidence reduction in 392 children under five years of age when LAI is deployed as a seasonal malaria prevention tool. We found that 393

if the protective efficacy of a new LAI decays immediately after injection, for example an exponential or 394

bi-phasic-like decay, then the LAI is unlikely to achieve non-inferiority over SMC-SP-AQ in current SMC 395 settings in a clinical trial. This exploration assumed non-inferiority criterion is required for testing, and we 396 only explored LAI half-life of protection in the range of 30 to 150 days. In contrast, when the protective 397 efficacy of LAIs is long-lasting and decays only after some time (i.e. a sigmoidal decay), there is a stronger 398 chance of achieving non-inferiority when the duration of protection is close to the transmission length. 399 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. LAIs in a larger range of incidence and transmission settings. If the half-life of protective efficacy of a LAI 402 approaches the length of transmission season, and initial efficacy is sufficiently high (depending on the 403 transmission intensity (Fig. 2) ), then the development of new LAI should prioritize optimizing operational 404 delivery factors to ensure reasonable coverage to be as good as or better in averting clinical cases than 405 current SMC-SP+AQ implementation. 406

The estimated impact of tool properties 407

In general, the duration of the half-life and its shape of decay are the most relevant tool properties for 408 incidence reduction. 

Our non-inferiority analysis (Fig. 3) highlights that the establishment of non-inferiority in clinical trial 437 stages is challenging due to not only the high protective ability of SMC-SP+AQ but also clinical trial 438

designs. This motivates an important discussion on the clinical development of LAIs under the use-case of 439 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. 

Replacing SMC-SP+AQ with LAIs will likely prove beneficial in reducing deployment costs due to fewer 455 deployment rounds. In the absence of information on costs of LAIs (costs of goods and supply chain), we 456 are unable to adequately assess economic considerations and thus our results assume that coverage is the 457 main driver of implementation cost. If LAIs are assumed to have a longer clearance half-life and therefore 458 higher protective efficacy for longer than SP+AQ, resources are freed up through decreased deployment 459 rounds within a season. These resources could be reallocated to increase the overall coverage in a single 460 round of LAI in the target population, including populations in remote places. Additionally, the overall 461 adherence to the blood-stage clearing co-administration of antimalarials could be increased, thereby 462 reducing the probability of emergence of resistance. However, if transmission intensity is very high, we 463

found that the optimization of protective efficacy half-life and deployment coverage is insufficient to 464 adequately protect the targeted population. Instead, it might be necessary to expand the deployment of LAIs 465 to multiple administration rounds within a transmission season. 466

The optimization of deployment coverage of LAIs to reach non-inferiority to SMC-SP+AQ where optimal 467 coverage cannot be met exposes an additional use case for LAIs. If external circumstances, such as the 468 current COVID-19 pandemic, prevent the regular implementation of SMC and bed-nets campaigns, 469

millions of children will experience an increased risk of malaria 50 , LAIs may alleviate this burden. Our 470 analysis can aid the identification of minimal LAI coverages necessary to achieve given population impacts 471

and prevent a resurgence in malaria cases. 472 between operational constraints and tool properties to narrow down beneficial implementation settings and 480 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint use cases without the need for expensive field studies. Not only does this approach offer the opportunity to 481 assess the potential population impact of new tools currently under development, but it also provides a 482 methodology to assess the potential clinical trial outcomes. It assists the evaluation of clinical trial scenarios 483 that might be considered over several different malaria transmission and health-system settings to 484 supporting thinking on appropriate population impact endpoints that are suitable to inform decision making. 485

This is particularly true for existing interventions with high efficacy for which the establishment of non-486 inferiority in clinical non -inferiority trials is problematic, due to the required large sample sizes 51 . Here, 487

our approach offers first insights into the outcomes of such trials and the additional possibility to develop 488 clinical trial analysis tools. 489

Furthermore, beyond the current scope of our study, as more information on likely costs of LAIs become 490 available and further certainty in implementation and cold-chain needed, this work can serve as a basis for 491 cost-effectiveness or economic analysis. LAIs are co-administered with different blood-stage clearing drugs (different efficacies and/or potential 507 properties e.g. transmission blocking) or are deployed with other interventions such as insecticide-treated 508 bed-nets. And lastly, we explored SMC or LAI replacement in only children under 5 years of age and in 509 settings similar to where SMC is currently deployed such as Mali and Senegal. Further analysis could be 510 undertaken to assess LAI as seasonal prevention in children under 10 years of age, however we expect 511 conclusions to be similar in regards tool properties and coverage requirements. Our results also only hold 512 for assessing LAI as replacing SMC; we did not explore use cases of deploying LAI in perennial or other 513 settings in which SMC is not yet deployed. Alternative clinical metrics would need to be explored as LAI 514

in these use-cases are not a replacement tools, rather new tools and non-inferiority trials are not relevant. 515

Although this study focuses on the use of LAIs in seasonal malaria transmission settings, our findings 516 regarding the importance of protective efficacy half-life do provide first insights for potential use of a LAIs 517 in perennial malaria transmission settings. The protective efficacy half-life of a LAI will most likely dictate 518 the number of applications to children within one year to ensure effective clinical case reduction. 519

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. formulation atovaquone of 32 days in humans 16 and 8.9-19.6 hours of P218 in first-in-human trials 57 are 531 again likely insufficient in the use cases explored in our study and emphasize the need for longer-lasting 532

formulations to be useful as SMC replacements when deployed only once per season. Further use may be 533 possible for multiple applications within a season. 534

Here, we provided the first quantitative evaluation of the TPPs of future LAIs for malaria as a seasonal 536 prevention tool in children. Simulation analysis of LAIs in real-life implementation settings revealed that 537 the ability of LAIs to prevent clinical cases in children is strongly dependent on the length of the malaria 538 transmission season and transmission intensity. We also found it is important to focus on improving the 539 protective efficacy duration (half-life) of LAIs in development, as the speed of protective efficacy decay is 540

a key driver of overall impact or the chance to meet non-inferiority criteria compared to SMC-SP+AQ. 541

However, if a reasonable duration is possible (longer half-life and sigmoidal decay that supports protection 542 close to the length of transmission season) then development should focus on increasing deployment 543 coverage to optimizing the LAIs chance of higher impact. This provides evidence for the potential trade-544 offs between tool properties and operational constraints as LAIs are developed and deployed. In general, 545 our findings support the need for a thorough and combined investigation of tool properties and use cases in 546 the future development of LAIs. This combined effort includes earlier modelling alongside clinical studies 547 to provide evidence of translation of impact at population levels before late stage clinical studies and 548 optimize the success of new malaria tools. Our research here provides an initial foundation to support 549 dialogue between stakeholders, scientists, and clinicians at each clinical development stage of novel anti-550 infective LAI's to reduce clinical malaria incidence. LAIs have the potential to be a game changer in 551 protecting vulnerable populations from malaria. Our analysis serves as a stepping stone for the refinement 552

of TPPs for LAIs, thereby assisting the target-oriented use-case of development and implementation of new 553

LAIs. 554 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. transmission settings were defined using a factorial design covering all possible combinations of discrete 581 health system and vector specifications. The parameters defining the efficacy and delivery profiles of LAIs 582

(highlighted in bold in the third column) were sampled within the defined parameter space using Latin 583

Hypercube Sampling and simulated for each combination of settings. The effective coverage E14 describes 584 the probability that effective malaria treatment will occur within a 14-day period since symptoms onset. 585

Additional information on simulated transmission intensity can be found in the Supplement (Fig. S1 and 586 Table S1 ). 587 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint protective efficacy. Estimated relationships between initial protective efficacy and efficacy half-life for 624 different incidence reduction criteria (40%, 60% and 80%, line style and color) and clinical incidence 625 settings (increasing color intensity indicates an initial clinical incidence measured in cases per person per 626

year in the target age group of 0.5, 1, 1.5, 2, and 2.8). Each line shows the minimum required LAI 627 characteristics to reach the desired health goal at a 100% LAI deployment coverage at clinical trial stage, 628 with all parameter combinations below a line failing to meet those requirements. The panels show the 629 parameter space of attainable incidence reductions within the specified constrained ranges of initial 630 protective efficacy and half-life for exponential LAIs (a, b) , bi-phasic LAIs (c, d) and sigmoidal LAIs (e, f) 631

in settings with a short (Senegal-like a, c, e) or long (Mali-like b, d, f) malaria season. The incidence 632 reduction was calculated by comparing the incidence over one transmission season after application of the 633 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. protective efficacy) and calculating the incidence reduction by comparison to the initial clinical incidence 637 measured in cases per person per year0.25-5y in the respective transmission intensity setting. shapes and length of transmission season are shown in the Supplementary Figures S4 and S5 . 666 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint and low access to treatment (E14=0.1). In the grey area, non-inferiority of LAIs could not be established for 674 any coverage. The light blue frames capture the tool characteristics where non-inferiority could be reached 675 with a LAI coverage under the reference SMC-SP+AQ coverage of 60%. Further results for additional 676 settings and decay shapes are provided in the supplement ( Fig. S8 and S9 ). The coloured dots represent 677 four illustrative LAI profiles for which the corresponding predicted relative differences in cases per person 678 per year0.25-5y (Eq. 5) are calculated in (b-e) five years after LAI introduction over all LAI coverages as 679 compared to SMC-SP+AQ at 60% coverage (vertical dotted line). The predicted positive increase in relative 680 difference in yearly clinical cases (above the dotted horizontal line) means more clinical cases are averted 681

with LAIs than with SMC-SP+AQ. It thus illustrates the benefit of increasing sigmoidal LAI-coverage 682 above the minimal required coverage to achieve non-inferiority (shown by the grey coloured area). Due to 683 the chosen margin of non-inferiority (here 5%, see Material and Methods), LAIs are non-inferior for a slight 684 negative relative difference in cases per person per year0.25-5y. In the light-blue area in (b), a LAI coverage 685 lower than the SMC-SP+AQ coverage is sufficient to establish non-inferiority. The corresponding analysis 686

for exponential LAIs can be found in Fig. S10 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint Malaria specification and intervention set-up .................................................. 31  837   2. Non-inferiority analysis ........................................................................................................... 34  838   3. Additional analysis results ............................................................ . Error! Bookmark not defined. 839 the clinical trial setting .............................................................. 35  840   5. Additional analysis results in the implementation setting ...................................................... 36  841   6. Parameterization of SMC-SP+AQ to clinical trial data .............................................................. 43  842  843 844 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. relationship between prevalence and incidence is displayed in the two simulated seasonal settings 847 (Senegal/short season and Mali/long season) and health system access settings (low and high access) (Table  848 1) in absence of any interventions. The clinical incidence defined as the events per person per year in the 849 target age-group (0.25-5 years of age) is shown for the corresponding mean prevalence over one year in the 850 intervention age group (PfPR0.25-5y) and in children between 2-10 years (PfPR2-10y). The dotted horizontal 851 lines mark the incidence settings, simulated for all downstream analyses and can be found in Table S1 . 852 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. parameterisation of the interventions is further specified in Table 1 . 874 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. ( 3 FGH ) Eq. S10 911

The confidence interval of the hazard ratio is given by N [PQ% AH: N±I.PU×WX(N)] . Non-inferiority is 912 established if the upper limit of the derived 95% confidence interval, CIhigh, of the hazard ratios between 913 SMC and LAI lies below the upper limit for non-inferiority g. 914

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. 926 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. represent the mean and the 95%-confidence bands (shaded area) capture the distribution of incidence 936 reduction across all sampled values. Increasing color intensity represents increasing initial cases per person 937 per year0.25-5y). These results hold true for high access to healthcare settings. The conversion of initial cases 938 per person per year0.25-5y to prevalence can be found in Table S1 . 939 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. represent the mean and the 95%-confidence bands (shaded area) capture the distribution of incidence 947 reduction across all sampled values. Increasing color intensity represents increasing initial clinical 948 incidence (cases per person per year0.25-5y). These results hold true for high access to healthcare settings. 949

The conversion of initial cases per person per year0.25-5y to prevalence can be found in Table S1 . 950 951 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. introduction of SMC-SP+AQ (initial cases per person per year0.25-5y). The incidence reduction0.25-5y was 956 calculated in the implementation scenario after one year of implementation. The grey lines indicate the 957 additional incidence reduction 0.25-5y) achieved by increasing SMC-SP+AQ coverage from 62% to 100 %. 958 959 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 5, 2021. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint Figure S9 : Estimated minimal LAI coverage required during implementation stages to achieve non-972 inferiority in a given setting. Heatmap of the estimated minimal coverage (colour) of LAIs at which non-973 inferiority to SMC-SP+AQ (assuming a fixed SMC coverage of 60% in each of the 3 or 4 rounds) is 974 achieved for different combinations of protective efficacy decay, initial protective efficacy and protective 975 efficacy half-life. The results are displayed for intervention scenarios with a high access to care (E14=0.5) 976

in the two seasonal settings. In the grey area, non-inferiority could not be established for any combination 977 of tool properties. The LAI coverage could not be optimized for high transmission settings (initial cases per 978 person per year0.25-5y=2.9) because they fail to sufficiently protect the targeted population from clinical 979 CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint malaria even at full deployment coverage. Therefore, optimisation of the LAI deployment coverage could 980 not be conducted (Fig. S4) . 981 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ; https://doi.org/10.1101/2021.07.05.21250483 doi: medRxiv preprint is increasing with the ongoing malaria transmission season. After the last SMC round, there is an increase 1061 of prevalence caused by still ongoing transmission. Simultaneously, the transmission intensity is already 1062 decreasing as can be seen in the control cohort prevalence. 1063 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 5, 2021. ;

The effect of malaria control on Plasmodium falciparum in Africa between 2000 691 and 2015

WHO Policy Recommendation: Seasonal Malaria Chemoprevention (SMC) for Plasmodium 693 falciparum malaria control in highly seasonal transmission areas of the Sahel sub-region in Africa

Randomized Noninferiority Trial of Dihydroartemisinin-Piperaquine Compared 696 with Sulfadoxine-Pyrimethamine plus Amodiaquine for Seasonal Malaria Chemoprevention in Burkina 697 Faso

Effectiveness of Seasonal Malaria Chemoprevention in Children under Ten Years 699 of Age in Senegal: A Stepped-Wedge Cluster-Randomised Trial

Evaluation of direct and indirect effects of seasonal malaria chemoprevention in Mali

Seasonal malaria chemoprevention: successes and 703 missed opportunities

Seasonal Malaria Chemoprevention Coverage Survey in Guinea

Measuring the impact of seasonal malaria chemoprevention as part of routine 707 malaria control in Kita, Mali

Selection of drug resistance-mediating Plasmodium falciparum genetic 712 polymorphisms by seasonal malaria chemoprevention in Burkina Faso

Contrasting Asymptomatic and Drug Resistance Gene Prevalence of Plasmodium 715 falciparum in Ghana: Implications on Seasonal Malaria Chemoprevention

Markers of sulfadoxine-pyrimethamine resistance in Eastern Democratic 717 Republic of Congo; implications for malaria chemoprevention

Seasonal malaria chemoprevention with sulfadoxine-pyrimethamine plus amodiaquine in 719 children: A field guide. (World Health Organization

Injectable anti-malarials revisited: discovery and development of new agents 721 to protect against malaria

Long-acting injectable atovaquone nanomedicines for malaria prophylaxis

Malarial dihydrofolate reductase as a paradigm for drug development against 725 a resistance-compromised target

A human monoclonal antibody prevents malaria infection by targeting a new 727 site of vulnerability on the parasite

A public antibody lineage that potently inhibits malaria infection by dual binding to 729 the circumsporozoite protein

A quantitative approach to guide development of novel disease 731 interventions, submitted

A model for natural immunity to asexual blood stages of Plasmodium falciparum 733 malaria in endemic areas

Public health impact and cost-effectiveness of the RTS,S/AS01 malaria vaccine: 735 a systematic comparison of predictions from four mathematical models

A periodically-forced mathematical model for the seasonal 738 dynamics of malaria in mosquitoes

A mathematical model for the dynamics of malaria in 740 mosquitoes feeding on a heterogeneous host population

An epidemiological model of the incidence of acute illness in Plasmodium 742 falciparum malaria

Relationships between host infectivity to mosquitoes and asexual 744 parasite density in Plasmodium falciparum

Modelling reactive case detection strategies for interrupting 746 transmission of Plasmodium falciparum malaria

Defining the relationship between infection prevalence and clinical incidence of 748 Plasmodium falciparum malaria

Implementation, coverage and equity of large-scale door-to-door delivery of 750 Seasonal Malaria Chemoprevention (SMC) to children under 10 in Senegal

INDEPTH Network: contributing to the data revolution

Country profiles

Large Sample Properties of Simulations Using Latin Hypercube Sampling

Dynamics of the antibody response to Plasmodium falciparum infection in 760 African children

The time-course of protection of the RTS,S 762 vaccine against malaria infections and clinical disease

Protective Efficacy of Intermittent Preventive Treatment of Malaria in Infants 764 (IPTi) Using Sulfadoxine-Pyrimethamine and Parasite Resistance

Efficacy and safety of intermittent preventive treatment with sulfadoxine-766 pyrimethamine for malaria in African infants: a pooled analysis of six randomised

Gaussian Processes in Machine Learning

Population Pharmacokinetic Properties of Sulfadoxine and Pyrimethamine: a 773 Pooled Analysis To Inform Optimal Dosing in African Children with Uncomplicated Malaria

Statistical methods to derive efficacy estimates of anti-malarials for uncomplicated 776 Plasmodium falciparum malaria: pitfalls and challenges

Heteroskedastic Gaussian Process Modeling and Design 778 under Replication

Global sensitivity indices for nonlinear mathematical models and their Monte Carlo 780 estimates

Large-scale delivery of seasonal malaria chemoprevention to children under 10 in 782 Senegal: an economic analysis

The Costs of Seasonal Malaria Chemoprevention (SMC) in the Sahel 784

Sub-Region of Africa: A Multi-Country Cost Analysis of the ACCESS-SMC project

Assessing drug efficacy against Plasmodium falciparum liver stages in vivo

DSM265 for Plasmodium falciparum chemoprophylaxis: a randomised, double 790 blinded, phase 1 trial with controlled human malaria infection

Seasonal intermittent preventive treatment with artesunate and sulfadoxine-792 pyrimethamine for prevention of malaria in Senegalese children: a randomised, placebo-controlled, double-793 blind trial

Identifying and combating the impacts of COVID-19 on malaria

Revisiting the design of phase 797 III clinical trials of antimalarial drugs for uncomplicated Plasmodium falciparum malaria

Translational pharmacokinetics and pharmacodynamics of monoclonal antibodies

Antibodies against Plasmodium falciparum malaria at the molecular 802 level

Strategies to extend plasma half-lives of recombinant antibodies

A Novel Investigational Fc-Modified Humanized Monoclonal Antibody

Has an Extended Half-Life in Healthy Adults

Tolerability, and Pharmacokinetics of MEDI4893, an Investigational, 809 Extended-Half-Life, Anti-Staphylococcus aureus Alpha-Toxin Human Monoclonal Antibody

First-in-human clinical trial to assess the safety, tolerability and 812 pharmacokinetics of P218, a novel candidate for malaria chemoprotection

Distribution of malaria exposure in endemic countries in Africa considering 815 country levels of effective treatment

in rural western Burkina Faso Zongo et 1007 al.(2015) 3 . Decay of protective efficacy of SPAQ in the field over time was extracted from Fig.3 in Zongo 1008 et al.(2015) 3 and used to parameterized the decay functions as specified in OpenMalaria using a least 1009 squares approach combined with a Gaussian-Process optimization. The trial described in Zongo et The OM implementation of the best parameter set for the intervention cohort is able to well capture the 1039 protective efficacy described in 3 . The protective efficacy (Figure S12)

Plasmodium falciparum malaria control in highly seasonal transmission areas of the Sahel sub-region in 1065 Africa

Statistical methods to derive efficacy estimates of anti-malarials for uncomplicated 1067 Plasmodium falciparum malaria: pitfalls and challenges

Randomized Noninferiority Trial of Dihydroartemisinin-Piperaquine Compared 1069 with Sulfadoxine-Pyrimethamine plus Amodiaquine for Seasonal Malaria Chemoprevention in Burkina 1070 Faso

Anopheles funestus (Diptera: Culicidae) in a humid savannah area of western 1072 Burkina Faso: bionomics, insecticide resistance status, and role in malaria transmission

Distribution of malaria exposure in endemic countries in Africa considering 1076 country levels of effective treatment

 Table S4 . 1011 

We used Latin-Hypercube sampling (LHS) to generate 5000 samples of the decay function parameters for 1014 protective efficacy (L, k, E0) within the parameter bounds given in Table S4 . The trial was simulated with 1015 these parameters and five seeds per parameter-set for the intervention and control cohort. The protective 1016 efficacy E of SMC -SP+AQ in the simulated trial compared to the simulated controls (without intervention) 1017was extracted by comparing cases per person (cpp) between the intervention group (cppint) and control 1018 group (cppcont) over the trial period as follows: 1019The residual sum of squares (RSS) between the protective efficacy given in 3 and 1021 protective efficacy from OM simulations was extracted and a Gaussian process (GP) regression was trained 1022to predict the RSS between trial results and OM simulation with the parameters of the protective efficacy 1023 decay. 1024The true over predicted RSS of the hold-out of 1000 data points is shown in Fig. S11 .