Overview

Treatment pathways for radical cure of P. vivax infections were incorporated into a previously developed transmission model to better understand the potential impact of introducing tafenoquine into case management practices on population-level transmission dynamics (Fig A in S1 Text) [26]. Quantitative G6PD diagnostic testing was implemented in the model using Gaussian mixture models for men and women fitted to data from studies reporting phenotypic G6PD activity, calibrated to G6PDd genotype prevalence from a large Brazilian survey (Fig D and Fig F in S1 Text). We accounted for misclassification with the SD Biosensor G6PD test as a reference (Fig B in S1 Text). We evaluated the differential impact of varying 8-aminoquinoline eligibility criteria and efficacy, and conducted sensitivity analyses on adherence and prevalence of low CYP2D6 metabolism. Different scenarios of tafenoquine rollout in 2021 were simulated for municipalities with active transmission to evaluate the public health impact up to the year 2030. The treatment scenarios to be modelled were specified in advance of epidemiological data acquisition.

P. vivax transmission model

We adapted a previously published mathematical model of P. vivax transmission to model case management with tafenoquine in Brazilian settings [26]. Fig A in S1 Text shows a compartmental representation of the transmission model. The model is implemented in 2 complementary formats: first, as a deterministic compartmental model described by a system of ordinary differential equations, and, second, as an individual-based stochastic model. Notably, the 2 model implementations provide identical predictions for large population sizes, when stochastic effects average out. The population dynamics of multiple mosquito species are described using compartmental models. The model was further adapted to reflect the epidemiology of P. vivax in Brazil by incorporating 2 distinct modes of transmission (peri-domestic and occupational) (see S1 Text for details). The model code is available online at https://github.com/MWhite-InstitutPasteur/Pvivax_TQ_IBM.

Parameter estimates for the Brazilian setting

We used data from 2018 case reports from the Malaria Epidemiological Surveillance Information System (SIVEP-Malaria) database to estimate the incidence (light-microscopy-positive slides or positive rapid diagnostic tests reported per 1,000 population) for each municipality and to extract the eligibility criteria for treatment among cases (age, pregnancy, and lactation status) [27]. The model was calibrated for 126 Brazilian municipalities reporting at least 100 cases from January to December 2018 using case notification data from SIVEP-Malaria for incidence and occupational exposure risk, the Brazilian Institute of Geography and Statistics (IBGE) databases for population size and mean age, and a large genotyping survey for G6PDd prevalence (Fig E, Fig F, and Table D in S1 Text). We assumed the reported incidence represented the majority of clinical cases in the population. Parameters for treatment efficacy, low CYP2D6 metabolism, and G6PDd diagnostics were assumed to be the same across the municipalities and were estimated from the literature (S1 Text). For the Brazilian setting, we assumed an average relapse rate of 1/69 days⁻¹ based on work by Daher et al. [28]. Model calibration details are provided in Table A in S1 Text. Ethical approval was not required for our study.

Since several municipalities cover a large spatial area and population, we included an additional peri-urban setting to model variations between urban and rural transmission. Since the municipality of Manaus accounts for a population of over 2 million residents, with a large malaria-free urban area, we accounted for the intra-municipality heterogeneity in transmission by only modelling the peri-urban area of Manaus, informed by data from a longitudinal cohort [29]. Reported aggregate results include only peri-urban Manaus.

Intervention scenarios

Using transmission levels from 2018, we compared a baseline scenario with the current radical cure regimen with chloroquine and primaquine to several scenarios for tafenoquine case management. For each setting, each scenario was simulated 100 times, and incidence across the time period was smoothed with a moving average. All results are reported as the median with minimum–maximum range or as a 95% uncertainty interval (UI) of values between the 2.5 and 97.5 percentiles. A population size of 100,000 was simulated to ensure constant transmission, and outputs were converted to the estimated 2018 population size. For settings with fewer than 100 reported cases, we assumed the median effect size from the simulated settings to estimate impact. Additional scenarios were simulated to account for uncertainty in parameter estimates for pre-existing adherence and assumptions of the impact of low CYP2D6 metabolism.

Malaria epidemiology in Brazil

Due to strong case management practices, Brazil has seen a significant decrease in malaria over the last 15 years, with its lowest level reported in 2016, at 151,000 cases [4]. The number of cases has since increased, reaching 237,000 reported cases in 2018, of which more than 90% were due to P. vivax. Including mixed infections, 217,000 P. vivax cases were reported, along with 21,000 cases of P. falciparum (Table 1). Approximately 60% of reported cases were in males, and approximately 64% in adults aged 16 years or older. Pregnant women represented 1.5% of all malaria cases.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of 2018 malaria case reports from the national surveillance system (SIVEP-Malaria).

https://doi.org/10.1371/journal.pmed.1003535.t001

The age and sex distributions of cases across different transmission settings showed high levels of heterogeneity (Fig 1A). In Itaituba municipality (an archetype of occupational exposure) in Pará state, working-aged males were the main driver of cases, with an incidence of 23 cases per 1,000. In São Gabriel da Cachoeira (an archetype of peri-domestic transmission), cases were observed across all ages and equally across sex, with a higher proportion in young children. A very high incidence of 267 cases per 1,000 was observed in this setting. In peri-urban Manaus, where transmission is significantly higher than in the municipality as a whole (114 cases versus 6 cases per 1,000), a combination of occupational exposure and exposure around the household likely drove transmission dynamics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Effective hypnozoite clearance with current radical cure with primaquine and with novel tafenoquine radical cure treatment pathways.

(A) The age and sex distribution of cases reported in 2018 across 3 example settings (left to right: Itaituba in Pará state [PA], peri-urban area of Manaus in Amazonas [AM], and São Gabriel da Cachoeira in Amazonas) shows the varying underlying P. vivax epidemiology. (B) In the treatment pathway for the current radical cure regimen with primaquine (PQ), age and sex determine eligibility for primaquine, while everyone receives chloroquine (CQ). Of those who receive PQ, some will not fully adhere to the full regimen. Of those who do fully adhere, a certain proportion will be subject to treatment failure due to low CYP2D6 metabolism and lack of drug efficacy for the given total dose. (C) We estimated that 43.1%, 42.9%, and 42.2% of cases in these 3 respective settings were effectively treated with primaquine. (D) The novel treatment pathway for radical cure with tafenoquine (TQ) introduces G6PD testing and a second age criterion for treatment eligibility. Along with infants and pregnant women, individuals with G6PD deficiency (≤30% G6PD activity) receive only CQ. Those with intermediate activity (30%–70%) and aged under 16 years are prescribed CQ and PQ. Those 16 years and older and with normal G6PD activity (≥70%) are prescribed CQ and TQ. (E) We estimated that with this novel pathway, radical cure with tafenoquine along with prescription for primaquine could achieve 59.0%, 60.5%, and 61.9% of effective radical cure rates among cases in Itaituba, Manaus, and São Gabriel da Cachoeira, respectively.

https://doi.org/10.1371/journal.pmed.1003535.g001

Estimating the effective radical cure rate with chloroquine and primaquine

The current radical cure treatment regimen in Brazil for P. vivax cases is chloroquine 25 mg/kg total dose over 3 days and primaquine 3.5 mg/kg total dose over 7 days. The proportion of individuals eligible for primaquine depends on the underlying demographics of the population. In Fig 1B, for the baseline scenario, we estimated that among primaquine eligible cases, 66.7% fully adhered to the 7-day regimen, 8.1% were low CYP2D6 metabolisers, and primaquine had 71.3% efficacy (S1 Text). Based on these assumptions, we estimate that a median 43.7% (range 19.4%, 43.7%) of reported cases of P. vivax and mixed infections in 2018 were effectively treated (defined as fully adherent, with normal CYP2D6 metabolism, and with 100% clearance of parasites) under the current radical cure regimen, with some variability across different settings (Fig 1C; Table E and Fig G in S1 Text). This corresponds to a total of 92,150 effectively treated P. vivax cases in 2018.

Novel chloroquine, primaquine, and tafenoquine treatment pathway

In the standard tafenoquine introduction scenario (Fig 1D), we assume the same primaquine adherence rate of 66.7% and a prevalence of low CYP2D6 metabolisers of 8.1%. Since tafenoquine is prescribed as a single dose, we assume 100% adherence. We assume that tafenoquine and primaquine have a comparable drug efficacy in the Brazilian population of 71.3%, since no significant difference was observed in the phase III clinical trial [11]. We assume no role of CYP2D6 in tafenoquine metabolism (S1 Text). Assuming the novel treatment pathway was introduced in the same epidemiological context of 2018, we estimate a median 59.8% radical cure rate (range 26.6%, 61.9%), resulting in an additional 16.1% of clinical cases being effectively treated, the equivalent of 34,950 additional cases (Table E and Fig I in S1 Text).

Impact of tafenoquine on transmission in archetype settings

With plans for rollout in the coming years, we simulated tafenoquine introduction in Brazil in 2021 assuming similar transmission levels to those in 2018. The previously described novel treatment pathway (scenario 1 [S1]) was compared to a baseline scenario with no tafenoquine introduction and scenarios 2–6 (S2–S6) with variable age eligibility criteria, tafenoquine efficacy, and pre-existing primaquine adherence (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Tafenoquine introduction scenarios.

https://doi.org/10.1371/journal.pmed.1003535.t002

Transmission intensity and the age and sex distribution of cases are major factors influencing the impact of tafenoquine. In Fig 2 and Table 3, we show the variable impact of different tafenoquine introduction scenarios (S1–S6) across the 3 archetype P. vivax transmission settings previously described in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Effect of introducing tafenoquine on P. vivax incidence in low-, moderate-, and high-transmission settings for scenarios 1–4.

(A, C, E) Model simulated P. vivax incidence under no intervention (baseline) and under scenarios 1–4 introducing tafenoquine with testing for G6PD deficiency in 2021 in Itaituba (A), peri-urban Manaus (C), and São Gabriel da Cachoeira (E). Incidence represents the average of 100 independent simulations per scenario, with a moving-average smoothing of the data. Incidence was calibrated to clinical cases reported in 2018 per 1,000 population. (B, D, F) The effect size (in percent) is reported as the percentage reduction in incidence after 5 years (incidence in 2025 compared to 2020) for scenarios 1–4 in Itaituba (B), peri-urban Manaus (D), and São Gabriel da Cachoeira (F). (G) The 5-year post-intervention effect size in all simulated municipalities. A Loess fitted line and 95% confidence interval bands (shading) are shown. For municipalities with an incidence below 25 cases per 1,000 population, stochastic fade out in the absence of importation prevents accurate assessment of impact.

https://doi.org/10.1371/journal.pmed.1003535.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Effect of introducing tafenoquine on P. vivax incidence in low-, moderate-, and high-transmission settings for scenarios 1–6.

https://doi.org/10.1371/journal.pmed.1003535.t003

With increasing P. vivax incidence, we expect a smaller effect size of tafenoquine on population-level transmission (Fig 2G). In higher transmission settings, individuals in the population are highly exposed to P. vivax parasites and develop higher levels of immunity; as a result, the majority of infections are subpatent, and only a small subset of cases will develop clinical symptoms and seek treatment (Fig J in S1 Text). For example, in Itaituba municipality in Pará state, the model predicts that 18% of infections result in clinical disease, while in São Gabriel da Cachoeira municipality in Amazonas, clinical cases represent only 8% of all infections. As a result, the impact of S1 in Itaituba is much higher than in São Gabriel da Cachoeira (31.9% versus 9.1%) since case management is able to reach a higher proportion of cases (Table 3).

The age and sex distribution of cases has a major impact on eligibility for tafenoquine and the potential impact on a population level. Introduction of a higher efficacy tafenoquine dose (S2) would improve effective radical cure rates and cause a greater reduction in transmission in highly occupational exposure settings like Itaituba, where the majority of cases are in adult males (Fig 2A and 2B). In moderate and highly peri-domestic exposure settings, where many cases are reported in those under 16 years old, the availability of a paediatric formulation of tafenoquine (S3) would be more advantageous than S2 (Fig 2C–2F). For S3 in São Gabriel da Cachoeira, a 21.2% reduction in transmission could result in 11,600 cases averted over 5 years if children are also treated, as compared to 4,400 cases averted when treating only adults (Table 3; Table F in S1 Text).

The level of impact of tafenoquine on transmission depends greatly on the assumption of the pre-existing primaquine adherence rate (Fig K in S1 Text). In a low-transmission setting such as Itaituba, if the pre-existing adherence rate is as low as 30% (S6), tafenoquine introduction will have a significant impact on transmission, reducing incidence from 23 cases to 9.1 per 1,000, with a 5-year effect size of 60.8%. In contrast, if the pre-existing primaquine adherence is 90%, tafenoquine only has a marginal impact, reducing transmission from 23 cases to 21.7 cases per 1,000, with a 4.4% effect size. In a high-transmission setting like São Gabriel da Cachoeira, where low adherence could be an issue due to a significant indigenous population, tafenoquine could reduce transmission by a predicted 16.4%. Even such a small reduction in transmission would result in an important reduction of clinical cases, averting potentially an additional 10,100 cases over a 5-year period (Table F in S1 Text).

National-level impact of tafenoquine

In the standard tafenoquine introduction scenario (S1), tafenoquine is expected to reduce P. vivax transmission regardless of the main exposure routes or transmission intensity (Fig 3) by increasing the mean effective hypnozoite clearance rate across simulated years 2021 to 2030 from a baseline of 42.3% (95% UI 40.5%–44.2%) to 62.1% (95% UI 54.2%–67.7%). By the end of the first year, we estimate a median effect size of 5.1% in transmission (Fig L and Table I in S1 Text). We expect an additional 19.9% increased impact in the second year, followed by an incremental increase until transmission stabilises over time, reaching a new equilibrium (Fig 3E). The effect size and cumulative number of averted cases will vary by setting and transmission intensity. Five years post-introduction, we estimate a median 37.5% reduction in transmission (Fig 3D). Ten low-transmission municipalities (incidence < 5 cases per 1,000) are predicted to reach over 80% reduction in transmission within 5 years post-introduction (Fig 3C). However, among moderate- and high-transmission settings (incidence > 25 cases per 1,000), we predict a median 18.2% effect size (Table J in S1 Text).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Effect of introducing tafenoquine across the P. vivax endemic region of Brazil.

(A) P. vivax incidence per 1,000 population across all municipalities (n = 424) reporting at least 1 case in 2018. (B) Comparison of incidence and total cases reported in 2018, with selected municipalities labelled (peri-urban Manaus setting shown as star). (C) The effect size (percent change in incidence) 5 years after introduction of the standard tafenoquine introduction scenario (scenario 1). (D) The median effect size across all municipalities over time for scenarios 1–6 (S1–S6). S1 is shown as a solid line and S2–S6 are shown as dotted lines. (E) The total cumulative number of averted cases reached 5 years post-introduction for S1 as compared to the baseline scenario for all 424 municipalities. (F) The total cumulative number of averted cases over time for all municipalities for S1–S6. Maps generated with shapefiles from the R package malariaAtlas.

https://doi.org/10.1371/journal.pmed.1003535.g003

While the majority of settings will not reach elimination with S1, 4 municipalities and peri-urban Manaus would observe over 8,000 averted cases over this period. An estimated 177,000 new P. vivax cases would be averted during the first 5 years of tafenoquine rollout among simulated settings (Fig 3F; Table H in S1 Text). Among settings reporting fewer than 100 cases in 2018 (i.e., settings that were not simulated), a 37.5% median effect size over 5 years could result in an additional 36,800 averted cases assuming no change in annual reported cases (Table H in S1 Text). Overall, over 214,000 P. vivax cases would be averted across all endemic settings in S1 (Table H in S1 Text).

We assessed the impact of alternative strategies, including a paediatric formulation of tafenoquine and a safely administered high-efficacy dose across all settings (S2–S4). Improving tafenoquine efficacy to 90% in adults (S2) would have a greater overall impact on transmission than introducing a paediatric formulation of the standard efficacy dose in children ages 2 to <16 (S3) (Fig 3D). Compared to S1, S2 would improve impact by an additional 14.0% within 5 years, as compared to 7.5% for S3, since larger majorities of settings have cases in adult males due to occupational exposure (Table 4). On the other hand, S3 would result in more cases averted due to a higher burden among children in high-transmission settings being able to benefit from radical cure with tafenoquine. Having a safe high-efficacy dose for all ages (S4) would have the most benefit, reducing transmission by an additional 22.1% compared to S1 (median effect size of 59.6%) and averting twice as many cases (Table 4). Such a scenario would help reach an 80% reduction in P. vivax transmission in 18 simulated municipalities (mainly those with low transmission) and reach elimination in 4 municipalities. The vast majority of municipalities across S1–S4 would not reach elimination with improved case management alone.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Effect of introducing tafenoquine across the P. vivax endemic region of Brazil.

https://doi.org/10.1371/journal.pmed.1003535.t004

Adherence to primaquine can be highly variable across different populations worldwide; therefore, we assessed the impact of different levels of pre-existing primaquine adherence on the effect of introducing tafenoquine. The 37.5% median impact of S1 relied heavily on the assumption of a 66.7% pre-existing primaquine adherence rate in the Brazilian population. If the pre-existing primaquine adherence rate is higher, at 90% (S5), tafenoquine only contributes a 17.7% reduction over 5 years. However, if the pre-existing adherence rate is as low as 30% (S6), tafenoquine introduction has a substantial impact on transmission, with a median effect size of 66.8% (Fig 3D; Table 4).

The total number of treatment courses of chloroquine, primaquine, and tafenoquine and the number of G6PD tests administered over a 5-year post-intervention period were estimated from model outputs (Table 4). As transmission decreases overtime, fewer chloroquine doses would need to be prescribed overall. At least 846,000 G6PD tests will need to be delivered over this period in S1. Even after the introduction of tafenoquine, more primaquine courses will be prescribed than tafenoquine courses due to the demography of the case population.

Other contributing factors

Another alternative strategy to improving radical cure among those ineligible for tafenoquine (children or those with intermediate G6PD activity) is switching from the low (3.5 mg/kg total dose) to high (5 mg/kg total dose) primaquine regimen. Switching to higher efficacy tafenoquine and primaquine (S7 in Table J in S1 Text) as compared to only high-efficacy tafenoquine (S2) would improve the 5-year median effect size by 4.0% (from 51.5% to 55.5%), especially in settings with more cases in children who would benefit from improved primaquine efficacy (Table J and Fig M in S1 Text).

Cytochrome P450 2D6 (CYP2D6) is an important enzyme in the activation of drugs in the liver and in particular of 8-aminoquinoline [17]. If low CYP2D6 metabolism also compromises tafenoquine efficacy (S8 in Table J in S1 Text), we predict a reduced impact (from 37.5% to 28.6%) (Table J and Fig N in S1 Text). A lower prevalence of low CYP2D6 metabolism of 4% (S9 in Table J in S1 Text), compared to 8.1%, would not have a significant impact. However, a higher prevalence of 20% (S10 in Table J in S1 Text) would compromise 8-aminoquinoline efficacy more significantly.