Skip to main content
eLife logoLink to eLife
. 2020 Oct 27;9:e59872. doi: 10.7554/eLife.59872

Sex-based differences in clearance of chronic Plasmodium falciparum infection

Jessica Briggs 1,, Noam Teyssier 1, Joaniter I Nankabirwa 2,3, John Rek 2, Prasanna Jagannathan 4, Emmanuel Arinaitwe 2,5, Teun Bousema 6,7, Chris Drakeley 7, Margaret Murray 4, Emily Crawford 8, Nicholas Hathaway 9, Sarah G Staedke 5, David Smith 10, Phillip J Rosenthal 1, Moses Kamya 2,3, Grant Dorsey 1, Isabel Rodriguez-Barraquer 1,, Bryan Greenhouse 1,
Editors: Urszula Krzych11, Eduardo Franco12
PMCID: PMC7591246  PMID: 33107430

Abstract

Multiple studies have reported a male bias in incidence and/or prevalence of malaria infection in males compared to females. To test the hypothesis that sex-based differences in host-parasite interactions affect the epidemiology of malaria, we intensively followed Plasmodium falciparum infections in a cohort in a malaria endemic area of eastern Uganda and estimated both force of infection (FOI) and rate of clearance using amplicon deep-sequencing. We found no evidence of differences in behavioral risk factors, incidence of malaria, or FOI by sex. In contrast, females cleared asymptomatic infections at a faster rate than males (hazard ratio [HR]=1.82, 95% CI 1.20 to 2.75 by clone and HR = 2.07, 95% CI 1.24 to 3.47 by infection event) in multivariate models adjusted for age, timing of infection onset, and parasite density. These findings implicate biological sex-based differences as an important factor in the host response to this globally important pathogen.

Research organism: P. falciparum

Introduction

Malaria, a protozoan infection of the red blood cells, remains one of the greatest global health challenges (World malaria report, 2019). Infection with malaria parasites results in a wide range of clinical disease presentations, from severe to uncomplicated; in addition, in hyperendemic areas, asymptomatic infections are common (Bousema et al., 2014). It is well established that chronic asymptomatic infection with Plasmodium falciparum, the most common and fatal malaria parasite, can lead to morbidity for those infected and contribute to ongoing transmission (Bousema et al., 2014; Okell et al., 2012; Tadesse et al., 2018; Slater et al., 2019). Characterization of these asymptomatic infections is paramount as they represent a major obstacle for malaria elimination efforts. Therefore, an understanding of how host immunity and parasite factors interact to cause disease tolerance is required. While age-specific immunity to malaria in hyperendemic areas is well-characterized, less attention has been paid to the possibility of a sex bias in malarial susceptibility despite evidence for a male bias in malaria infections in non-human animals and a male bias in the prevalence of other human parasitic infections (Zuk and McKean, 1996; Klein, 2004; Roberts et al., 2001).

The clearest evidence for sexual dimorphism in malaria susceptibility is in pregnant women, who are at greater risk of malaria infection and also experience more severe disease and higher mortality (Desai et al., 2007). However, multiple studies performed in different contexts have demonstrated a male bias in incidence and/or prevalence of malaria infection in school-aged children and adults (Molineaux et al., 1980; Pathak et al., 2012; Landgraf et al., 1994; Camargo et al., 1996; Abdalla et al., 2007); this bias is more well established in hypoendemic areas, but has also been observed in hyperendemic regions (Houngbedji et al., 2015; Mulu et al., 2013). Where there is a male bias in malaria infection, it has often been postulated that these differences in malaria incidence or prevalence stem from an increased risk of males acquiring infection due to socio-behavioral factors (Pathak et al., 2012; Camargo et al., 1996; Moon and Cho, 2001; Finda et al., 2019). However, because biological sex itself has been demonstrated to affect responses to other pathogens, an alternative hypothesis is that the sexes may have different responses to the malaria parasite once infected (Nhamoyebonde and Leslie, 2014; Bernin and Lotter, 2014; Fischer et al., 2015; Fish, 2008).

Estimating the host response to P. falciparum infection requires close follow-up of infected individuals, sensitive detection of parasites, and the ability to distinguish superinfection, which is common in endemic areas, from persistent infection. To test the hypothesis that sex-based differences in host–parasite interactions affect the epidemiology of malaria, we intensively followed a representative cohort of individuals living in a malaria endemic area of eastern Uganda. Using frequent sampling, ultrasensitive quantitative PCR (qPCR), and amplicon deep-sequencing to genotype parasite clones, we were able to accurately detect the onset of new infections and follow all infections over time to estimate their duration. Using these data, we show that females cleared their asymptomatic infections more rapidly than males, implicating biological sex-based differences as important in the host response to this globally important pathogen.

Results

Cohort participants and P. falciparum infections

This analysis involved data from 477 children and adults (233 males and 244 females) that were followed for a total of 669.6 person-years (Table 1). 25 participants, 10 males and 15 females, were enrolled after initial enrollment (Figure 1); median duration of follow-up in those who were dynamically enrolled was 0.84 years (IQR 0.69–1.23) compared to 1.45 years (IQR 1.43–1.47) in those enrolled during initial enrollment. 149 of 477 participants (31.2%) included in the analysis had at least one P. falciparum infection detected (Figure 1). 114 participants had 822 successfully genotyped samples and had infections characterized by clone and by infection event. 35 samples (from 35 unique participants) had very low-density infections (<1 parasite/µL) that could not be genotyped; these infections were characterized at the infection event level only. We achieved a read count of >10,000 for 92% of genotyped samples, identifying 45 unique AMA-1 clones in our population (frequencies and sequences in Supplementary file 1e). At the clone level, the proportion of baseline infections out of all infections in males was 117/171 (68.4%), compared to 68/116 (58.8%) baseline infections in females (p=0.10). At the infection event level, there were 54/104 (51.9%) baseline infections in males and 45/89 (50.6%) baseline infections in females (p=0.89).

Table 1. Behavioral risk factors for malaria infection and measures of malaria burden in study population, stratified by age and sex.

Metric Age and gender categories
All <5 years old 5–15 years old 16 years and older
Male Female Male Female Male Female Male Female
Number of participants, n 233 244 73 84 101 71 59 89
Median days of follow-up per participant 530.0 530.0 525.0 524.5 530.0 531.0 530.0 530.0
Slept under LLIN the previous night 53.6% 56.3% 54.1% 56.3% 47.9% 47.8% 63.7% 64.3%
Person-years of follow-up 324.2 345.4 87.3 96.7 152.6 118.5 84.22 130.1
Number of overnight trips 44 107 21 19 9 10 14 78
Incidence of overnight trips*, (95% CI) 0.14
(0.09–0.20)
0.31
(0.20–0.49)
0.24
(0.14–0.42)
0.20
(0.09–0.42)
0.06
(0.03–0.13)
0.08
(0.03–0.24)
0.17
(0.09–0.31)
0.60
(0.30–1.18)
Episodes of malaria* 11 13 5 2 5 9 1 2
Incidence of malaria**, (95% CI) 0.03
(0.02–0.06)
0.04
(0.02–0.09)
0.06
(0.02—0.16)
0.02
(0.00–0.11)
0.03
(0.01–0.08)
0.08
(0.03–0.23)
0.01
(0.00–0.08)
0.02
(0.00–0.17)
Number of routine visits, n 4316 4583 1164 1293 2034 1568 1118 1722
Prevalence of microscopic parasitemia*** 2.9% 1.4% 1.8% 1.1% 4.4% 2.7% 1.3% 0.5%
Prevalence of parasitemia by qPCR 14.4% 9.2% 5.8% 3.7% 17.0% 15.3% 18.5% 7.7%
Geometric mean parasite density**** 3.41 3.06 4.86 13.06 6.31 4.20 1.09 1.02
Median complexity of infection, (IQR) 3 (1–7) 2 (1–2) 1 (1–2.5) 2 (1–2.3) 4 (2–9) 2 (1–2) 1 (1–2) 1 (1–2)

*Malaria includes one episode (female,<5 years old), due to non-falciparum species (P. malariae).

**per person-year.

***Parasitemia by light microscopy includes one episode (female, 5–15 years old) due to non-falciparum species (P. ovale).

****Geometric mean parasite density in parasites/µL of all qPCR-positive routine visits.

Figure 1. Study design.

Figure 1.

Behavioral malaria risk factors and measures of malaria burden

There was no difference in reported rates of LLIN use the previous night by sex (Table 1). Women over the age of 16 traveled overnight outside of the study area more than men (incidence rate ratio [IRR] for females vs. males = 3.61, 95% confidence interval [CI] 1.83 to 7.13), a potential risk factor for malaria exposure. Antimalarial use outside the study clinic was reported only four times (3 females and one male, all under the age of 5). In this region receiving regular rounds of IRS, the incidence of symptomatic malaria was low in all age categories, and there was no evidence of a difference in incidence of symptomatic malaria by sex overall (IRR for females vs. males = 1.11, 95% CI 0.49 to 2.52) or when adjusted for age (IRR = 1.26, 95% CI 0.54 to 2.96). In contrast, prevalence ratio (PR) of P. falciparum parasitemia by microscopy in females versus males across all age categories was 0.49 (95% CI 0.36 to 0.65), with relative differences in prevalence most pronounced in the oldest age group. Similar findings were seen when prevalence was assessed by ultrasensitive qPCR, with PR = 0.64 in females vs. males (95% CI 0.43 to 0.96), again with the largest differences seen in the oldest age group. Adjusting for age as a categorical variable, LLIN use, and travel did not qualitatively change prevalence ratios for microscopic parasitemia (PR in females vs. males = 0.57, 95% CI 0.42 to 0.77) or for qPCR-positive parasitemia (PR in females vs. males = 0.67, 95% CI 0.60 to 0.76). We also found no evidence for a difference in parasite density as determined by qPCR between males and females after adjusting for age as a continuous variable (p=0.47). Median complexity of infection (COI) was higher in males than in females overall, driven primarily by a higher COI in male school-aged children.

Force of infection by age and sex

To determine whether higher infection prevalence in males was due to an increased rate of infection, we used longitudinal genotyping to calculate the force of infection (FOI, number of new blood stage infections per unit time). Overall, the FOI was low, with new infections occurring on average less than once every 5 years (Table 2). There was no evidence for a significant difference in FOI by sex overall (IRR for females vs. males = 0.88, 95% CI 0.48 to 1.62 by clone and IRR = 0.83, 95% CI 0.52 to 1.33 by infection event). There was also no evidence for a significant difference in FOI by sex when adjusted for age category (IRR = 0.88, 95% CI 0.47 to 1.63 by clone and IRR = 0.83, 95% CI 0.53 to 1.31 by infection event). For analysis both by clone and by infection event, there was a trend toward higher FOI in males compared to females. We performed a sensitivity analysis by decreasing the number of skips necessary to declare an infection cleared (Supplementary file 1f). Performing the same analysis using one skip for clearance instead of 3 skips increased the FOI in all groups and increased the trend toward higher FOI in males, with IRR for females vs. males when adjusted for age category = 0.71, 95% CI 0.41 to 1.24 by clone and IRR = 0.75, 95% CI 0.47 to 1.22 by infection event.

Table 2. Molecular force of infection (FOI) by clone and by infection event, stratified by age and sex.

Molecular force of infection (FOI) Sex Age category
All <5 years 5–15 years 16 years or older
By clone, ppy* (95% CI) All 0.18 (0.13–0.24) 0.14 (0.07–0.28) 0.19 (0.08–0.43) 0.20 (0.08–0.46)
Male 0.19 (0.12–0.30) 0.16 (0.07–0.39) 0.19 (0.10–0.37) 0.22 (0.09–0.54)
Female 0.17 (0.09–0.31) 0.12 (0.03–0.49) 0.19 (0.08–0.45) 0.18 (0.06–0.54)
By event, ppy* (95% CI) All 0.14 (0.11–0.18) 0.09 (0.06–0.16) 0.16 (0.09–0.30) 0.16 (0.08–0.32)
Male 0.16 (0.11–0.22) 0.13 (0.07–0.27) 0.18 (0.11–0.28) 0.15 (0.07–0.29)
Female 0.13 (0.08–0.21) 0.06 (0.02–0.18) 0.14 (0.07–0.27) 0.17 (0.07–0.41)

*per person-year.

Rate of clearance of infection and duration of infection by sex

Since females had a lower prevalence of infection but similar rate of acquiring infections compared to males, we evaluated whether there was a difference between sexes in the rate at which infections were cleared. Asymptomatic infections were included in this analysis if they were not censored as stated in the methods. At the clone level, 105 baseline infections and 53 new infections were included; there was a slightly higher proportion of baseline infections in males (68/99, 68.7%), compared to the proportion of baseline infections in females (37/59, 62.7%) (p=0.49). At the infection event level, 58 baseline infections and 51 new infections were included and there was no difference in the proportion of baseline infections by sex, with 32/60 (53.3%) baseline infections in males and 26/49 (53.1%) baseline infections in females (p=1.0).

Unadjusted hazard ratios for clearing infecting clones showed that asymptomatic infections cleared naturally (i.e., when not treated by antimalarials) at nearly twice the rate in females vs. males (hazard ratio (HR) 1.92, 95% CI 1.19 to 3.11, Table 3). In addition, new infections cleared faster than baseline infections and monoclonal infections cleared faster than polyclonal infections. Unadjusted hazard ratios for clearance of infection events (as opposed to clones) also showed faster clearance in females vs. males (HR = 2.30, 95% CI 1.20 to 4.42). Results were similar in multivariate models including age, gender, the period during which the infection was first observed, and parasite density, demonstrating faster clearance in females vs. males (HR = 1.82, 95% CI 1.20 to 2.75 by clone and HR = 2.07, 95% CI 1.24 to 3.47 by infection event). Complexity of infection was not included in the final adjusted model because the model fit the data less well when COI was included as a predictor. In both adjusted models, new infections cleared faster than baseline infections. Higher parasite densities were associated with slower clearance by clone and by infection event, but the effect size was larger when data were analyzed by infection event (HR = 0.44, 95% CI 0.35 to 0.54). There was no evidence for interaction between age and sex in either adjusted model. We performed a sensitivity analysis by decreasing the number of skips necessary to declare an infection cleared. Regardless of whether three skips, two skips, or one skip was used to determine infection clearance, females cleared their infections faster than males both by clone and by infection event (Supplementary file 1g).

Table 3. Hazard ratios for rates of clearance of infection, by clone and by infection event.

Predictors Categories Hazard ratio by clone (95% CI) Hazard ratio by infection event (95% CI)
Unadjusted Adjusted Unadjusted Adjusted
Sex Male ref ref ref ref
Female 1.92 (1.19–3.11) 1.82 (1.20–2.75) 2.30 (1.20–4.42) 2.07 (1.24–3.47)
Age 16 years or greater ref ref ref ref
5–15 years 0.66 (0.39–1.10) 0.81 (0.49–1.36) 0.82 (0.39–1.74) 1.27 (0.72–2.25)
<5 years 1.64 (0.79–3.41) 1.55 (0.76–3.17) 2.01 (0.80–5.00) 1.75 (0.87–3.53)
Complexity of infection (COI) Polyclonal (COI > 1) ref -- ref --
Monoclonal (COI = 1) 1.63 (1.03–2.57) -- 0.95 (0.38–2.34) --
Infection status Present at baseline ref ref ref ref
New infection 1.94 (1.22–3.07) 1.75 (1.05–2.94) 4.66 (2.58–8.42) 4.32 (2.59–7.20)
Parasite density * 0.85 (0.69–1.06) 0.81 (0.65–1.00) 0.41 (0.32–0.51) 0.44 (0.35–0.54)

*Increasing parasite density (log10) in parasites/microliter, as measured by qPCR.

We next estimated durations of asymptomatic infection by age and sex using results from a model that included these covariates (Figure 2). Durations of infection ranged from 103 days to 447 days by clone, and from 87 to 536 days by infection event. Males had a longer duration of infection across all age categories. Children aged 5–15 years had the longest duration of infection, followed by adults. Therefore, overall, males aged 5–15 years had the longest estimated duration of infection by either clone (447 days) or infection event (526 days).

Figure 2. Estimates of duration of infection from sex- and age-adjusted model.

Figure 2.

Estimated duration of infection in days, calculated by adjusting the point estimate of the baseline hazard by the coefficients of the sex- and age-adjusted model. Error bars represent standard errors of duration obtained from variance in the model coefficients. Point estimates of duration are labeled (*).

Discussion

Previous studies have reported a higher prevalence of malaria infection in males compared to females, with the difference often ascribed to differences in exposure (Molineaux et al., 1980; Pathak et al., 2012; Landgraf et al., 1994; Camargo et al., 1996; Abdalla et al., 2007; Houngbedji et al., 2015; Mulu et al., 2013). In a cohort study in eastern Uganda, we noted higher prevalence of malaria infection in males compared to females. By closely following a cohort of children and adults and genotyping every detected infection with sensitive amplicon deep-sequencing, we were able to estimate both the rate of infection (FOI) and duration of infection and to compare these measures by sex. We found that lower prevalence in females did not appear to be due to lower rates of infection but rather due to faster clearance of asymptomatic infections. To our knowledge, this is the first study to report a sex-based difference in the duration of malaria infection.

Though there are some conflicting reports in the literature, the majority of studies of malaria incidence and/or prevalence that evaluated associations with sex in late childhood, adolescence and adulthood have found a male bias in the observed measure of burden (Molineaux et al., 1980; Pathak et al., 2012; Landgraf et al., 1994; Camargo et al., 1996; Abdalla et al., 2007; Houngbedji et al., 2015; Mulu et al., 2013). We note that this is more often observed in hypoendemic settings and may be confounded by factors such as treatment-seeking behavior; however, this male bias has been reported in studies of both P. vivax and P. falciparum. Overall, these studies consistently suggest that males exhibit higher incidence and/or prevalence of malaria that begins during late childhood, persisting through puberty and the majority of adulthood (excepting the years when pregnancy puts women at higher risk). One possible explanation put forward for the sex-specific difference in burden has been that males are more frequently bitten by malaria-carrying mosquitos due to behavioral differences such as working outside, not sleeping under a net, or traveling for work (Pathak et al., 2012; Camargo et al., 1996; Moon and Cho, 2001; Finda et al., 2019). In our study, however, there were no statistically significant differences in malaria incidence or FOI by sex, though there was a trend toward higher FOI in males. We also saw no evidence of behavioral trends that would result in more infections in males; in fact, older women in our study did most of the traveling outside the study area (an area of low transmission compared to surrounding areas). We did not assess work habits as part of our study questionnaire, but the fact that we observe similar patterns in prevalence by sex in all age categories makes this a less likely explanation. Therefore, in this particular cohort, it appears that the observed male bias in parasite prevalence is best explained by slower clearance of infection in males. We cannot rule out that males were differentially exposed in the recent past (e.g., had a higher force of infection when transmission was higher, which could have affected immunity), and we plan to genotype samples from a prior cohort from 2011 to 2017 to test this hypothesis.

Very few studies have been conducted to explore immunological differences between males and females in their response to the malaria parasite. RTS,S vaccination is associated with higher all-cause mortality in girls compared to boys, and a trend toward higher risk of fatal malaria has been noted in vaccinated girls compared to boys, suggesting possible sexual dimorphism in immunological responses to malaria (Klein et al., 2016). The comprehensive Garki project found a male bias in the prevalence of P. falciparum infection after the age of 5 years that was noted to increase after control measures including IRS and mass drug administration. They also found that females had higher levels of certain antibodies against P. falciparum compared to males, but saw no difference in these levels after control measures (Molineaux et al., 1980). Hormonal differences have been posited to play a role in these sex-based immunological differences to malaria infection; for example, studies in mice show that testosterone appears to downregulate the immune response to malaria (Delić et al., 2011; Wunderlich et al., 1991). One study of irradiated sporozoite vaccination in mice showed that female mice vaccinated after puberty were better protected than males following parasite challenge, but this sex difference was not seen if the mice were vaccinated prior to puberty, suggesting a role for sex hormones in the development of immunity (Vom Steeg et al., 2019). Testosterone was again associated with decreased protection from malaria in that study (Vom Steeg et al., 2019). In addition, in Kenya, two studies showed that dehydroepiandrosterone sulfate (DHEAS) levels were significantly associated with decreased parasite density in both males and females, even after adjustment for age, but neither study directly compared males to females (Kurtis et al., 2001; Leenstra et al., 2003). Given that levels and types of sex hormones differ between males and females, as does the age of onset of puberty, the interaction between these hormones, age, and protection against malaria is likely to be complicated. We did not detect an interaction between age and sex in our model. This could be due to a lack of statistical power or imperfect detectability, as adults have lower density infections than children due to improved anti-parasite immunity (Bousema et al., 2014; Okell et al., 2012). In addition, the adult age group was comprised of a wide age range, which included post-menopausal women and might have blunted the effect of sex in that age group if there is a hormonal basis for some of this effect. Repeating our analysis using different age categories (<8 years, 8–13 years, and 13+ years) did not significantly change our findings. It is also possible that sex-based differences are explained by a different mechanism, such as differences in innate immunity, Y-linked chromosomal factors in males, increased X-linked gene expression in females, or by a combination of factors. More studies are needed to elucidate the relationship between sex-based biological differences between males and females and their impact on the development of effective antimalarial immunity in humans.

Of the variables we evaluated in addition to sex, baseline infection status and parasite density were most strongly associated with the rate of clearance of infection. Infections that were already present at the beginning of the study and persisted past the left censoring date may have been a non-random selection of well-established asymptomatic infections that were present at baseline at a higher frequency than average because they had a fundamentally different trajectory than newly established asymptomatic infections. Higher parasite densities were also associated with slower clearance of infection in both adjusted models, but the effect was most pronounced when the data were analyzed by infection event. This may be because low-density infections that we were unable to successfully genotype were only included in the infection event analysis, and these events tended to have short durations. The inclusion of parasite density in our multivariate models did not meaningfully alter associations between sex and duration of infection, providing evidence that the sex-based differences in duration were not mediated primarily by differences in parasite density in our cohort.

A limitation of our study is the statistical model’s assumption that all infections clear at the same rate, which was necessary because we did not observe the beginning of most infections. To rely less on this assumption, we adjusted for baseline infections, allowing them to have a different rate of clearance than new infections; this did not change our primary finding of a difference in rate of clearance by sex. We also acknowledge there may be other unmeasured confounders, such as genetic hemoglobinopathies or sex-based differences in unreported outside antimalarial use, that could affect our results. Another caveat is that our findings may not be generalizable to areas with different transmission intensity given that our study was conducted in a very specific setting: an area with previously very high transmission intensity that has been greatly reduced in recent years by repeated rounds of IRS.

Additionally, because it is difficult to genotype low-density infections and because parasite densities fluctuate over time, we allowed several ‘skips’ in detection before declaring an infection cleared or the same clone in an individual a new infection with the same clone. This requires the assumption that re-infection with the same clone is relatively rare, which is reasonable given the high genetic diversity and low force of infection seen in this setting; this assumption has been made in other longitudinal genotyping studies (Felger et al., 2012; Smith et al., 1999). Our method may have biased us toward longer durations of infection and fewer new infections overall, but this is unlikely to have introduced any significant bias in the observed associations by sex. We performed a sensitivity analysis to ensure that changing the number of skips allowed did not significantly change the main finding regarding longer duration of infection in males and our findings were consistent. Given decreasing parasite densities over time and a setting with low force of infection and low EIR, assuming perfect recovery of clones is likely to artificially inflate the force of infection in this cohort. There were not enough infections to perform a rigorous analysis of the distribution of clones within and between households, but given that the overall force of infection was quite low, the probability of re-infection with the same clone already present in a participant from another member in the household (which could bias toward longer duration of infection) was low and unlikely to have introduced any significant bias by sex.

In summary, we estimated the clearance of asymptomatic P. falciparum infections by genotyping longitudinal samples from a cohort in Nagongera, Uganda, using sensitive amplicon deep-sequencing and found that females cleared their infections at a faster rate than males; this finding remained consistent when adjusting for age, baseline infection status, and parasite density. Furthermore, we found no conclusive evidence for a sex-based difference in exposure to infection, either behaviorally or by FOI. Though there have long been observed differences in malaria burden between the sexes, there is still little known about sex-based biological differences that may mediate immunity to malaria. Unfortunately, much reported malaria data is still sex-disaggregated. Our findings should encourage epidemiologists to better characterize sexual dimorphism in malaria and motivate increased research into biological explanations for sex-based differences in the host response to the malaria parasite.

Materials and methods

Study setting and population-level malaria control interventions

This cohort study was carried out in Nagongera sub-county, Tororo district, eastern Uganda, an area with historically high malaria transmission. However, 7 rounds of indoor residual spraying (IRS) from 2014 to 2019 have resulted in a significant decline in the burden of malaria (Nankabirwa et al., 2019). Pre-IRS, the daily human biting rate (HBR) was 34.3 and the annual entomological inoculation rate (EIR) was 238; after 5 years of IRS, in 2019 the daily HBR was 2.07 and overall annual EIR was 0.43 as reported by Nankabirwa et. al (Nankabirwa et al., 2020).

Study design, enrollment, and follow-up

All members of 80 randomly selected households with at least two children were enrolled in October 2017 using a list generated by enumerating and mapping all households in Nagongera sub-county (Nankabirwa et al., 2020). The cohort was dynamic such that residents joining the household were enrolled and residents leaving the household were withdrawn. Data for this analysis was collected from October 1st, 2017 through March 31st, 2019; participants were included if they had at least 6 months of contiguous follow-up. Data from the 25 dynamically enrolled participants contributed to all analyses, including that of baseline infections. Participants were followed at a designated study clinic open daily from 8 AM to 5 PM. Participants were encouraged to seek all medical care at the study clinic and avoid the use of antimalarial medications outside of the study. Routine visits were conducted every 28 days and included a standardized clinical evaluation, assessment of overnight travel outside of Nagongera sub-county, and collection of blood by phlebotomy for detection of malaria parasites by microscopy and molecular studies. Participants came in for non-routine visits in the setting of illness. Blood smears were performed at enrollment, at all routine visits and at non-routine visits if the participant presented with fever or history of fever in the previous 24 hr. Participants with fever (>38.0°C tympanic) or history of fever in the previous 24 hr had a thick blood smear read urgently. If the smear was positive, the patient was diagnosed with malaria and treated with artemether-lumefantrine. Participants with asymptomatic parasitemia as detected by qPCR or microscopy were not treated with antimalarials, consistent with Uganda national guidelines. Study participants were visited at home every 2 weeks to assess use of long-lasting insecticidal nets (LLINs) the previous night.

Laboratory methods

Thick blood smears were stained with 2% Giemsa for 30 min and evaluated for the presence of asexual parasites and gametocytes. Parasite densities were calculated by counting the number of asexual parasites per 200 leukocytes (or per 500, if the count was less than 10 parasites per 200 leukocytes), assuming a leukocyte count of 8,000/μL. A thick blood smear was considered negative if examination of 100 high power fields revealed no asexual parasites. For quality control, all slides were read by a second microscopist, and a third reviewer settled any discrepant readings. In our experienced microscopists’ hands, the lower limit of detection is approximately 20–50 parasites/µL.

For qPCR and genotyping, we collected 200 µL of blood at enrollment, at each routine visit, and at the time of malaria diagnosis. DNA was extracted using the PureLink Genomic DNA Mini Kit (Invitrogen) and parasitemia was quantified using an ultrasensitive varATS qPCR assay with a lower limit of detection of 0.05 parasites/µL (Hofmann et al., 2015). Samples with a parasite density >= 0.1 parasites/µL blood were genotyped via amplicon deep-sequencing. All samples positive for asexual parasites by microscopy but negative for P. falciparum by qPCR were tested for the presence of non-falciparum species using nested PCR (Snounou et al., 1993).

Sequencing library preparation

Hemi-nested PCR was used to amplify a 236 base-pair segment of apical membrane antigen 1 (AMA-1) using a published protocol (Miller et al., 2017), with modifications (Supplementary file 1a). Samples were amplified in duplicate, indexed, pooled, and purified by bead cleaning. Sequencing was performed on an Illumina MiSeq platform (250 bp paired-end).

Bioinformatics methods

Data extraction, processing, and haplotype clustering were performed using SeekDeep (Hathaway et al., 2018), followed by additional filtering (Lerch et al., 2019). Supplementary file 1b shows the full bioinformatics workflow.

Data analysis

A clone was defined as a genetically identical group of parasites, for example with identical haplotypes. Because polyclonal infections can occur due to co-infection (one mosquito bite transmitting multiple clones) or superinfection (multiple bites), we analyzed the infection data both by clone and by infection event. For analysis by clone, each unique clone was counted as an infection and each clone’s disappearance as a clearance event. For analysis by infection event, any new clones seen within 3 visits of the date of the first newly detected clone(s) were grouped together and considered one new ‘infection event.’ Clearance of infection for these events required that all clones in the group be absent. A baseline infection was defined as a clone or infection event (group of clones) detected in the first 60 days of observation of a participant. New infections were defined as a new clone or infection event (group of clones) detected in a participant after day 60.

Imperfect detection of P. falciparum clones is known to be a limitation of all PCR-based genotyping methods; both biological factors, such as deep organ sequestration of clones in the lifecycle of the parasite, and methodological factors, such as difficulty amplifying minority clones (especially at lower parasite densities), results in imperfect detectability of all clones present in a single blood sample drawn on a single day (Felger et al., 2012; Koepfli and Mueller, 2017; Nguyen et al., 2018; Koepfli et al., 2011; Ross et al., 2012). Though amplicon deep-sequencing is less likely than other methods to miss minority clones (Lerch et al., 2017), detectability was a particular concern in our study, in which the majority of infections were submicroscopic and parasite densities declined over time [Supplementary file 1c]. To account for the fact that a clone might be missed in a single sample due to fluctuations in parasite density and/or methodological limitations, we allowed 3 ‘skips’ in detection and classified an infection as cleared only if it was not identified in four contiguous samples from routine visits. If the infection (as determined by clone or by infection event) was absent for four routine samples in a row, the last date of detection would be the end date of that infection. Therefore, for a clone that had previously infected an individual to be classified as a new infection, it had to have been absent from that individual for at least four routine visits. Additional details are found in Supplementary file 1d. This decision was supported by the fact that the diversity of the genotyped AMA-1 amplicon was quite high (expected heterozygosity = 0.949), resulting in a low probability (5.1%) for infection with the same clone by chance.

Data analysis was conducted in R (R Development Core Team, 2019) and Python (Python Language Reference, 2020). Travel was self-reported at routine visits, and LLIN use on the previous night was reported by field entomology teams who surveyed households every two weeks. Microscopic parasite prevalence was defined by the number of smear-positive routine visits over all routine visits. Parasite prevalence by qPCR was defined as the number of qPCR-positive visits over all routine visits. Thus, these measures represent the average prevalence during routine visits. Prevalence ratios were computed using Poisson regression with generalized estimating equations to adjust for repeated measures. Comparison of parasite density by sex was made using linear regression with generalized estimating equations to adjust for repeated measures. Force of infection (FOI) was defined as the number of new infections, including malaria episodes, divided by person time. Since the start of baseline infections was not observed, baseline infections were not included in calculating FOI. Poisson regression with generalized estimating equations to account for repeated measures was used to estimate malaria incidence, FOI, and to calculate incident rate ratios (IRR) for malaria and FOI. Hazards for clearance of untreated, asymptomatic infections were estimated using time-to-event models (shared frailty models fit using R package ‘frailtypack,’ version 2.12.2) (Rondeau and Gonzalez, 2005; Rondeau et al., 2012). Infections were censored for this analysis if they were only observed in the first three months or the last three months (before January 01, 2018 and after January 01, 2019) because they were not observed for long enough to determine whether clearance occurred. If an infection was observed for only one timepoint, it was assigned a duration of 14 days. These models assumed a constant hazard of clearance and included random effects to account for repeated measures in individuals. Parasite density was included in the model as a time-varying covariate. Duration of infection in days was calculated as 1/adjusted hazard.

Data accessibility

Data from the cohort study is available through an open-access clinical epidemiology database resource, ClinEpiDB at https://clinepidb.org/ce/app/record/dataset/DS_51b40fe2e2. Genotyping data and code used to generate tables and figures is available on GitHub (Briggs, 2020; copy archived at swh:1:rev:cf6c3256e609f4f136fc8d90f9cae1d61d6d8d63).

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Jessica Briggs, Email: Jessica.Briggs@ucsf.edu.

Urszula Krzych, Walter Reed Army Institute of Research, United States.

Eduardo Franco, McGill University, Canada.

Funding Information

This paper was supported by the following grants:

  • National Institute of Allergy and Infectious Diseases T32 AI007641-16 to Jessica Briggs.

  • National Institute of Allergy and Infectious Diseases U19AI089674 to Grant Dorsey.

  • Fogarty International Center K43TW010365 to Joaniter I Nankabirwa.

  • Fogarty International Center D43TW010526 to Emmanuel Arinaitwe.

Additional information

Competing interests

Reviewing editor, eLife.

No competing interests declared.

Author contributions

Conceptualization, Data curation, Formal analysis, Methodology, Writing - original draft, Writing - review and editing.

Formal analysis, Visualization, Methodology.

Resources, Supervision, Project administration.

Supervision, Project administration.

Conceptualization, Supervision, Methodology, Writing - review and editing.

Supervision, Project administration.

Supervision, Funding acquisition, Writing - review and editing.

Resources, Supervision, Funding acquisition, Writing - review and editing.

Conceptualization, Writing - review and editing.

Resources, Supervision.

Software, Methodology, Writing - review and editing.

Resources, Supervision, Writing - review and editing.

Conceptualization, Writing - review and editing.

Supervision, Funding acquisition, Writing - review and editing.

Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing.

Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing.

Conceptualization, Formal analysis, Supervision, Validation, Methodology, Writing - review and editing.

Conceptualization, Formal analysis, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration, Writing - review and editing.

Ethics

Human subjects: IRB approval for the PRISM2 cohort study was obtained in Uganda (UNCST HS-119ES, SOMREC: 2017-099), UK (LSHTM: 14266), and US (UCSF IRB: 17-22544, Stanford: UCSF IRB reliance). Informed consent was obtained from all participants prior to enrollment in the study per IRB guidelines.

Additional files

Supplementary file 1. (a) AMA-1 hemi-nested PCR protocol for amplicon deep-sequencing. (b) Bioinformatics workflow. (c) Declining qPCR density over time in the cohort. (d) Detailed explanation of skip rule criteria. (e) Haplotype sequences and frequencies. (f) Sensitivity analysis of molecular force of infection: Table 2 replicated using 2 skips or 1 skip. (g) Sensitivity analysis of duration of infection: Table 3 replicated using 2 skips or 1 skip.
elife-59872-supp1.docx (293.7KB, docx)
Transparent reporting form

Data availability

Data from the PRISM2 cohort study is available through a novel open-access clinical epidemiology database resource here: https://clinepidb.org/ce/app/record/dataset/DS_51b40fe2e2. Sequencing data is available on Github at https://github.com/EPPIcenter/sex_based_differences as referenced in the paper. In addition. all sequences of haplotypes are included in Supplementary file 1.

The following dataset was generated:

Dorsey G. 2020. PRISM2 ICEMR Cohort. ClinEpiDB. DS_51b40fe2e2

References

  1. Abdalla SI, Malik EM, Ali KM. The burden of malaria in Sudan: incidence, mortality and disability--adjusted life--years. Malaria Journal. 2007;6:97. doi: 10.1186/1475-2875-6-97. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bernin H, Lotter H. Sex Bias in the outcome of human tropical infectious diseases: influence of steroid hormones. Journal of Infectious Diseases. 2014;209 Suppl 3:S107–S113. doi: 10.1093/infdis/jit610. [DOI] [PubMed] [Google Scholar]
  3. Bousema T, Okell L, Felger I, Drakeley C. Asymptomatic malaria infections: detectability, transmissibility and public health relevance. Nature Reviews Microbiology. 2014;12:833–840. doi: 10.1038/nrmicro3364. [DOI] [PubMed] [Google Scholar]
  4. Briggs J. sex_based_differences. jbriggs7Github. 2020 https://github.com/EPPIcenter/sex_based_differences
  5. Camargo LM, dal Colletto GM, Ferreira MU, Gurgel SM, Escobar AL, Marques A, Krieger H, Camargo EP, da Silva LH. Hypoendemic malaria in Rondonia (Brazil, western Amazon region): seasonal variation and risk groups in an urban locality. The American Journal of Tropical Medicine and Hygiene. 1996;55:32–38. doi: 10.4269/ajtmh.1996.55.32. [DOI] [PubMed] [Google Scholar]
  6. Delić D, Ellinger-Ziegelbauer H, Vohr HW, Dkhil M, Al-Quraishy S, Wunderlich F. Testosterone response of hepatic gene expression in female mice having acquired testosterone-unresponsive immunity to Plasmodium chabaudi malaria. Steroids. 2011;76:1204–1212. doi: 10.1016/j.steroids.2011.05.013. [DOI] [PubMed] [Google Scholar]
  7. Desai M, ter Kuile FO, Nosten F, McGready R, Asamoa K, Brabin B, Newman RD. Epidemiology and burden of malaria in pregnancy. The Lancet Infectious Diseases. 2007;7:93–104. doi: 10.1016/S1473-3099(07)70021-X. [DOI] [PubMed] [Google Scholar]
  8. Felger I, Maire M, Bretscher MT, Falk N, Tiaden A, Sama W, Beck HP, Owusu-Agyei S, Smith TA. The dynamics of natural Plasmodium falciparum infections. PLOS ONE. 2012;7:e45542. doi: 10.1371/journal.pone.0045542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Finda MF, Moshi IR, Monroe A, Limwagu AJ, Nyoni AP, Swai JK, Ngowo HS, Minja EG, Toe LP, Kaindoa EW, Coetzee M, Manderson L, Okumu FO. Linking human behaviours and malaria vector biting risk in south-eastern Tanzania. PLOS ONE. 2019;14:e0217414. doi: 10.1371/journal.pone.0217414. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Fischer J, Jung N, Robinson N, Lehmann C. Sex differences in immune responses to infectious diseases. Infection. 2015;43:399–403. doi: 10.1007/s15010-015-0791-9. [DOI] [PubMed] [Google Scholar]
  11. Fish EN. The X-files in immunity: sex-based differences predispose immune responses. Nature Reviews Immunology. 2008;8:737–744. doi: 10.1038/nri2394. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Hathaway NJ, Parobek CM, Juliano JJ, Bailey JA. SeekDeep: single-base resolution de novo clustering for amplicon deep sequencing. Nucleic Acids Research. 2018;46:e21. doi: 10.1093/nar/gkx1201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hofmann N, Mwingira F, Shekalaghe S, Robinson LJ, Mueller I, Felger I. Ultra-sensitive detection of Plasmodium falciparum by amplification of multi-copy subtelomeric targets. PLOS Medicine. 2015;12:e1001788. doi: 10.1371/journal.pmed.1001788. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Houngbedji CA, N'Dri PB, Hürlimann E, Yapi RB, Silué KD, Soro G, Koudou BG, Acka CA, Assi SB, Vounatsou P, N'Goran EK, Fantodji A, Utzinger J, Raso G. Disparities of Plasmodium falciparum infection, malaria-related morbidity and access to malaria prevention and treatment among school-aged children: a national cross-sectional survey in côte d'ivoire. Malaria Journal. 2015;14:7. doi: 10.1186/1475-2875-14-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Klein SL. Hormonal and immunological mechanisms mediating sex differences in parasite infection. Parasite Immunology. 2004;26:247–264. doi: 10.1111/j.0141-9838.2004.00710.x. [DOI] [PubMed] [Google Scholar]
  16. Klein SL, Shann F, Moss WJ, Benn CS, Aaby P. RTS,S malaria vaccine and increased mortality in girls. mBio. 2016;7:e00514-16. doi: 10.1128/mBio.00514-16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Koepfli C, Schoepflin S, Bretscher M, Lin E, Kiniboro B, Zimmerman PA, Siba P, Smith TA, Mueller I, Felger I. How much remains undetected? probability of molecular detection of human plasmodia in the field. PLOS ONE. 2011;6:e19010. doi: 10.1371/journal.pone.0019010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Koepfli C, Mueller I. Malaria epidemiology at the clone level. Trends in Parasitology. 2017;33:974–985. doi: 10.1016/j.pt.2017.08.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kurtis JD, Mtalib R, Onyango FK, Duffy PE. Human resistance to Plasmodium falciparum increases during puberty and is predicted by dehydroepiandrosterone sulfate levels. Infection and Immunity. 2001;69:123–128. doi: 10.1128/IAI.69.1.123-128.2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Landgraf B, Kollaritsch H, Wiedermann G, Wernsdorfer WH. Parasite density of Plasmodium falciparum malaria in Ghanaian schoolchildren: evidence for influence of sex hormones? Transactions of the Royal Society of Tropical Medicine and Hygiene. 1994;88:73–74. doi: 10.1016/0035-9203(94)90505-3. [DOI] [PubMed] [Google Scholar]
  21. Leenstra T, ter Kuile FO, Kariuki SK, Nixon CP, Oloo AJ, Kager PA, Kurtis JD. Dehydroepiandrosterone sulfate levels associated with decreased malaria parasite density and increased hemoglobin concentration in pubertal girls from western Kenya. The Journal of Infectious Diseases. 2003;188:297–304. doi: 10.1086/376508. [DOI] [PubMed] [Google Scholar]
  22. Lerch A, Koepfli C, Hofmann NE, Messerli C, Wilcox S, Kattenberg JH, Betuela I, O'Connor L, Mueller I, Felger I. Development of amplicon deep sequencing markers and data analysis pipeline for genotyping multi-clonal malaria infections. BMC Genomics. 2017;18:864. doi: 10.1186/s12864-017-4260-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Lerch A, Koepfli C, Hofmann NE, Kattenberg JH, Rosanas-Urgell A, Betuela I, Mueller I, Felger I. Longitudinal tracking and quantification of individual Plasmodium falciparum clones in complex infections. Scientific Reports. 2019;9:7. doi: 10.1038/s41598-019-39656-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Miller RH, Hathaway NJ, Kharabora O, Mwandagalirwa K, Tshefu A, Meshnick SR, Taylor SM, Juliano JJ, Stewart VA, Bailey JA. A deep sequencing approach to estimate Plasmodium falciparum complexity of infection (COI) and explore apical membrane antigen 1 diversity. Malaria Journal. 2017;16:490. doi: 10.1186/s12936-017-2137-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Molineaux L, Gramiccia G, Organization WH. The garki project : research on the epidemiology and control of malaria in the Sudan savanna of west africa. [February 18, 2020];World Health Organization. 1980 https://apps.who.int/iris/handle/10665/40316
  26. Moon JJ, Cho SY. Incidence patterns of vivax malaria in civilians residing in a high-risk county of Kyonggi-do (province), Republic of korea. The Korean Journal of Parasitology. 2001;39:293–299. doi: 10.3347/kjp.2001.39.4.293. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Mulu A, Legesse M, Erko B, Belyhun Y, Nugussie D, Shimelis T, Kassu A, Elias D, Moges B. Epidemiological and clinical correlates of malaria-helminth co-infections in southern Ethiopia. Malaria Journal. 2013;12:227. doi: 10.1186/1475-2875-12-227. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Nankabirwa JI, Briggs J, Rek J, Arinaitwe E, Nayebare P, Katrak S, Staedke SG, Rosenthal PJ, Rodriguez-Barraquer I, Kamya MR, Dorsey G, Greenhouse B. Persistent parasitemia despite dramatic reduction in malaria incidence after 3 rounds of indoor residual spraying in Tororo, Uganda. The Journal of Infectious Diseases. 2019;219:1104–1111. doi: 10.1093/infdis/jiy628. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Nankabirwa JI, Arinaitwe E, Rek J, Kilama M, Kizza T, Staedke SG, Rosenthal PJ, Rodriguez-Barraquer I, Briggs J, Greenhouse B, Bousema T, Drakeley C, Roos DS, Tomko SS, Smith DL, Kamya MR, Dorsey G. Malaria transmission, infection, and disease following sustained indoor residual spraying of insecticide in Tororo, Uganda. The American Journal of Tropical Medicine and Hygiene. 2020;103:1525–1533. doi: 10.4269/ajtmh.20-0250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Nguyen TN, von Seidlein L, Nguyen TV, Truong PN, Hung SD, Pham HT, Nguyen TU, Le TD, Dao VH, Mukaka M, Day NP, White NJ, Dondorp AM, Thwaites GE, Hien TT. The persistence and oscillations of submicroscopic Plasmodium falciparum and Plasmodium vivax infections over time in Vietnam: an open cohort study. The Lancet Infectious Diseases. 2018;18:565–572. doi: 10.1016/S1473-3099(18)30046-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Nhamoyebonde S, Leslie A. Biological differences between the sexes and susceptibility to tuberculosis. Journal of Infectious Diseases. 2014;209 Suppl 3:S100–S106. doi: 10.1093/infdis/jiu147. [DOI] [PubMed] [Google Scholar]
  32. Okell LC, Bousema T, Griffin JT, Ouédraogo AL, Ghani AC, Drakeley CJ. Factors determining the occurrence of submicroscopic malaria infections and their relevance for control. Nature Communications. 2012;3:1–9. doi: 10.1038/ncomms2241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Pathak S, Rege M, Gogtay NJ, Aigal U, Sharma SK, Valecha N, Bhanot G, Kshirsagar NA, Sharma S. Age-dependent sex Bias in clinical malarial disease in Hypoendemic regions. PLOS ONE. 2012;7:e35592. doi: 10.1371/journal.pone.0035592. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Python Language Reference . Python Software Foundation; 2020. http://www.python.org [Google Scholar]
  35. R Development Core Team . Vienna, Austria: R Foundation for Statistical Computing; 2019. https://www.R-project.org [Google Scholar]
  36. Roberts CW, Walker W, Alexander J. Sex-associated hormones and immunity to protozoan parasites. Clinical Microbiology Reviews. 2001;14:476–488. doi: 10.1128/CMR.14.3.476-488.2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Rondeau V, Mazroui Y, Gonzalez JR. Frailtypack : An R Package for the Analysis of Correlated Survival Data with Frailty Models Using Penalized Likelihood Estimation or Parametrical Estimation. Journal of Statistical Software. 2012;47:1–28. doi: 10.18637/jss.v047.i04. [DOI] [Google Scholar]
  38. Rondeau V, Gonzalez JR. Frailtypack: a computer program for the analysis of correlated failure time data using penalized likelihood estimation. Computer Methods and Programs in Biomedicine. 2005;80:154–164. doi: 10.1016/j.cmpb.2005.06.010. [DOI] [PubMed] [Google Scholar]
  39. Ross A, Koepfli C, Li X, Schoepflin S, Siba P, Mueller I, Felger I, Smith T. Estimating the numbers of malaria infections in blood samples using high-resolution genotyping data. PLOS ONE. 2012;7:e42496. doi: 10.1371/journal.pone.0042496. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Slater HC, Ross A, Felger I, Hofmann NE, Robinson L, Cook J, Gonçalves BP, Björkman A, Ouedraogo AL, Morris U, Msellem M, Koepfli C, Mueller I, Tadesse F, Gadisa E, Das S, Domingo G, Kapulu M, Midega J, Owusu-Agyei S, Nabet C, Piarroux R, Doumbo O, Doumbo SN, Koram K, Lucchi N, Udhayakumar V, Mosha J, Tiono A, Chandramohan D, Gosling R, Mwingira F, Sauerwein R, Paul R, Riley EM, White NJ, Nosten F, Imwong M, Bousema T, Drakeley C, Okell LC. The temporal dynamics and infectiousness of subpatent Plasmodium falciparum infections in relation to parasite density. Nature Communications. 2019;10:1433. doi: 10.1038/s41467-019-09441-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Smith T, Felger I, Fraser-Hurt N, Beck HP. Effect of insecticide-treated bed nets on the dynamics of multiple Plasmodium falciparum infections. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1999;93 Suppl 1:53–57. doi: 10.1016/S0035-9203(99)90328-0. [DOI] [PubMed] [Google Scholar]
  42. Snounou G, Viriyakosol S, Zhu XP, Jarra W, Pinheiro L, do Rosario VE, Thaithong S, Brown KN. High sensitivity of detection of human malaria parasites by the use of nested polymerase chain reaction. Molecular and Biochemical Parasitology. 1993;61:315–320. doi: 10.1016/0166-6851(93)90077-B. [DOI] [PubMed] [Google Scholar]
  43. Tadesse FG, Slater HC, Chali W, Teelen K, Lanke K, Belachew M, Menberu T, Shumie G, Shitaye G, Okell LC, Graumans W, van Gemert GJ, Kedir S, Tesfaye A, Belachew F, Abebe W, Mamo H, Sauerwein R, Balcha T, Aseffa A, Yewhalaw D, Gadisa E, Drakeley C, Bousema T. The relative contribution of symptomatic and asymptomatic plasmodium vivax and Plasmodium falciparum infections to the infectious reservoir in a Low-Endemic setting in Ethiopia. Clinical Infectious Diseases. 2018;66:1883–1891. doi: 10.1093/cid/cix1123. [DOI] [PubMed] [Google Scholar]
  44. Vom Steeg LG, Flores-Garcia Y, Zavala F, Klein SL. Irradiated sporozoite vaccination induces sex-specific immune responses and protection against malaria in mice. Vaccine. 2019;37:4468–4476. doi: 10.1016/j.vaccine.2019.06.075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. World malaria report . World Malaria Report 2019. WHO; 2019. https://www.who.int/publications-detail-redirect/9789241565721 [Google Scholar]
  46. Wunderlich F, Marinovski P, Benten WP, Schmitt-Wrede HP, Mossmann H. Testosterone and other gonadal factor(s) restrict the efficacy of genes controlling resistance to Plasmodium chabaudi malaria. Parasite Immunology. 1991;13:357–367. doi: 10.1111/j.1365-3024.1991.tb00289.x. [DOI] [PubMed] [Google Scholar]
  47. Zuk M, McKean KA. Sex differences in parasite infections: patterns and processes. International Journal for Parasitology. 1996;26:1009–1024. doi: 10.1016/S0020-7519(96)80001-4. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Urszula Krzych1
Reviewed by: Richard Price2

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Studies on sex-based differences in immune responses, in general, have established that females show enhanced activity in comparison to males. Recently, several studies have confirmed a male bias in the prevalence of human parasitic infections. The current study explores the possibility of sex dimorphism in susceptibility to Plasmodium malaria. On the basis of results from longitudinal studies conducted in Uganda, females cleared their malaria infections at a faster rate than males. These observations provide an impetus for increased research into biological explanations for sex-based differences in the host response to the malaria parasite.

Decision letter after peer review:

Thank you for submitting your article "Sex-based differences in clearance of chronic Plasmodium falciparum infection" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Richard Price.

As is customary in eLife, the reviewers have discussed their critiques with one another. What follows below is the Reviewing Editor's edited compilation of the essential and ancillary points provided by reviewers in their critiques and in their interaction post-review. Please submit a revised version that addresses these concerns directly. Although we expect that you will address these comments in your response letter we also need to see the corresponding revision in the text of the manuscript. Some of the reviewers' comments may seem to be simple queries or challenges that do not prompt revisions to the text. Please keep in mind, however, that readers may have the same perspective as the reviewers. Therefore, it is essential that you attempt to amend or expand the text to clarify the narrative accordingly.

As the editors have judged that your manuscript is of interest, but as described below additional work is required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

The reviewers considered these findings suggesting a bias in Plasmodium falciparum prevalence in males versus females to be very interesting and potentially important to a wide scientific audience. As the authors conclude, the sexual differences are caused by higher prevalence of infection in males because of longer infection carriage or slower clearance of infection. The convincing demonstration of true sexual dimorphism (as opposed to gender-related behavioral/occupational differences) in malaria risk would be an important scientific advance. Despite strengths of the study, the reviewers expressed several critical concerns which include lack of data, robustness of statistical analysis and some basic methodology. These important issues preclude formation of strong conclusions that males and females differed in malaria risk for intrinsic biological reasons versus external factors.

Essential revisions:

A general question for this study is the scope and generalizability of the problem the authors are addressing; while many earlier studies reported malaria epidemiology, only a handful reported sexual dimorphism in risk. The papers cited by the authors describe different (and inconsistent) evidence for sexual dimorphism. Pathak et al., 2012, cited in the paper states: "Sexual dimorphism does not exist in hyperendemic regions for both Plasmodium falciparum and P. vivax infections, although some reports note an increased parasite density in pubertal and post-pubertal males".

Issues with the age categorization: why were these age bands chosen? If hormonal differences play a role in gender differences then surely this would increase after puberty. Perhaps it did and that is why higher difference is emerging in the 5-15 group. The three categories lose important data on the age sex distributions that could be highly relevant to the analysis. Surely 5-12yrs would have helped explore the puberty issue. Another approach would be to retain age as a continuous variable. Either way this doesn't explain the smaller difference in adults – although there were low numbers of males and few events. Did the age limit extend into older/menopausal women and might this have contributed to the findings? Please discuss these observations in the conclusions. The discussion also needs to acknowledge the limitations of the study and how assumptions and residual confounding may have impacted upon the findings.

Swimmers plots or similar figure may be helpful to display duration of individual infections over time. These could be stratified by sex, would show the number and durations of "infections" and of "clones", and individuals could be sorted by age so that the relevant age groups could be indicated.

Please include information if baseline infections differed by sex. Baseline infections have significantly longer duration than new infections (which is counter-intuitive to this reviewer) and constituted a majority of the infections in their analyses. Did the authors see similar patterns of sexual dimorphism in their baseline infections as well as their new infections?

Data on important confounders are missing, for example, genetics/ hemoglobinopathies, drug use (by report and by drug levels during "asymptomatic" infections), time since residual spraying if done.

Terms used in the paper have overlapping meanings ("clone", "infection", "parasitemia", "malaria infection", "malaria", "asymptomatic malaria", "symptomatic malaria", etc) and not all are clearly defined-would be good to define each and use consistently. The definitions for "clearance" require some clarification.

Other major concerns surround the robustness of the conclusions based on the statistical analysis:

1) No direct statistical test is given to show that prevalence differs between the sexes after adjusting for behavioural factors (and age etc).

2) The analysis of infection clearance and duration of infection (and even the FOI) are all based on the arbitrary assumption that 4 negative routine visits are needed before an infection is deemed to have been cleared. How many of the results change when this assumption is relaxed?

3) The analysis of FOI between sexes may be confounded by what is classed as a "new infection". Does this result still hold when the definition of new infection is altered?

These are suggestions for validating the robustness of the conclusions.

1) The authors do not directly test whether sex is still a predictor of higher prevalence after other factors are considered, which one would expect to see in the typical analysis in order to make the claim that behavioural factors do not explain the difference between men and women. Instead, the authors argue by looking at a table of the mean characteristics of the population what is and is not different between men and women (Table 1). Their conclusion on the basis of a lack of significance between behavioural pattern beteween sexes this is not a statistical argument. If the authors wish to show that, after accounting for behavioural differences, there is still a link between sex and prevalence, they need a full statistical model that includes all factors. The authors should run a logistic regression looking at the predictors of infection status in individuals, and showing that sex is a significant predictor. Covariates in this model should be sex, age, the behavioural data (e.g. bednet usage, travel, etc), time point. The model should also include random-effects to account for repeated measures within individuals. If no other factors are included, it could be assumed that sex would be a predictor of infection. However, forwards and/or backwards regression could then be used to ask the question "is sex a significant predictor of infection once adjusting for other behavioural factors are included in the model?" Similar analysis is performed on the clearance data so this should be easily achievable for the authors. If sex is not a predictor of infection after adjusting for age and behavioural factors, then one might conclude that risk of detecting an infection is well explained by behavioural factors and not sex differences. If sex is still required to explain the data then that would be evidence for a difference in prevalence that is sex based on not behavioural.

2) The authors argue that women have faster clearance times of asymptotic infection but this analysis is directly dependent on the definition of when an infection is cleared. Follow-up is every 28 days in this study and the end of an infection was defined as a clonotype being absent for 4 routine visits (i.e. someone might have a clone present at one visit, and at a visit 112 days later, but not detected at all in-between and it would be considered the same infection). If one is reinfected with the same clone (which might be very probably in an environment where 10 of the clones detected account for 55% of all observed clones) it is possible to count a new infection as one long infection. This issue may both artificially prolonging clearance times in the analysis and reduce the number of "new infections" (i.e. lower the FOI calculated) and may bias observations for men and women. This demand for 4 consecutive negative visits before an infection is deemed to have been cleared is an arbitrary choice and to be believable the conclusions of the analysis should not depend strongly on this assumption. For this reason, the authors must check that varying this criteria doesn't alter the results too strongly.

a) The authors should explore how this definition of clearance (being 4 negative routine visits) impacts the conclusions. What if only one negative time point was used as the definition of the end of an infection? How would this affect the FOI for men and women, and how would it impact the clearance times observed? Since it is not possible to really know when an infection ended, and a new infection began, it is important to test that the conclusions are robust to these assumptions.

b) It may be that behaviour patterns impact the likelihood of being infected with the same clonotype twice. For example, if one group does more travel than another group, then they may be more likely to be infected with different clonotypes. This diversity of clonotypes would make an individual look like they had shorter durations of infection compared with a group that travels less and is re-infected over and over with the same clonotype. i.e. Women who travelled more may only look like they have shorter infection because they are not re-infected with the same clone as much as the men who travelled less. To test this, the authors should look at the distribution of clonotype frequencies and diversity between men and women (e.g. do women tend to have infection with a greater diversity in clonotypes than men?). This might also make men look like they have a lower FOI than they really do, and women a higher FOI than they really do, and it may make it seem like the two groups have the same FOI when actually they are very different.

c) The issue of 4 negatives being required for "clearance" may explain why the infection present in individuals at enrollment is longer than "new infection". This result would change if the end of an infection was defined as the first negative rather than requiring 4 negatives. This may indicate a bias in the analysis towards counting things as one infection that are actually reinfection with the same clonotype.

3) The authors should confirm that there is no difference in FOI between the sexes. Even though not significant, there appears to be a trend towards higher FOI in men compared with women in all three age groups (Table 2). This is important because it is not clear how different the FOI in men and women would need to be in order for it to manifest in the differences in prevalence that exist between men and women. It may be that there is really a difference in prevalence observed and that this study does not have statistical power to detect such a difference. Also, calculating FOI as the authors have done requires knowing "new infections" as opposed to just the carriage of old infection, and the definition of what is a new infection is going to influence these conclusions. The authors should further validate this conclusion from their data by: (a) changing their definition of new infection (as in point 2) and recalculating the FOI and determining if there is any significant difference between the sexes. (b) performing a time-to-next-infection analysis (survival analysis, e.g. cox regression), rather than calculate the FOI as new infections over person time. The latter would involve looking at the time until the next new clone appears in each person in a survival model (with appropriate censoring). The question would then be "is sex a significant covariate in a survival model of time to next infection (after including other behavioural factors as well, such as travel)?". If these two other approaches also reveal no difference in FOI between the sexes, then the conclusion that there is no difference in infection pressure on the sexes would be more robust/convincing.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Sex-based differences in clearance of chronic Plasmodium falciparum infection" for further consideration by eLife. Your revised article has been evaluated by a Senior Editor and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The information on baseline infections, a potential confounder to these analyses, still needs further clarification.

Subsection “Cohort participants and P. falciparum infections”: "25 participants.… were enrolled after initial enrollment…" Please state whether their baseline data contributed to analyses of baseline infections. It seems so from subsection “Data analysis” but should be stated explicitly.

Subsection “Cohort participants and P. falciparum infections” paragraph two: "35 samples had very low density infections (< 1 parasite/μL) that could not be genotyped and had infections characterized at the event level only". Did these come from 35 unique individuals who had a single infection? And do these represent the difference between 149 individuals with infections versus 114 individuals who had 822 typable samples? If so, please state this explicitly. It appears so from Figure 1.

"117/185.… baseline infections in males". Please state the proportion of all infections included in the analysis that were baseline infections, for males and females separately, at the clone and the infection event level. Baseline infections are an important confounder in this study because they differ from other infections and disproportionately occur in males at the clone level.

Subsection “Behavioral malaria risk factors and measures of malaria burden”: "differences in prevalence…" Prevalence is defined in subsection “Data analysis”, but it might be helpful to use the term "period prevalence" or a simple explanation here, since many non-epidemiology readers may assume that prevalence here refers to a point prevalence.

In the same section please define COI.

Subsection “Force of infection by age and sex”: "Force of infection…" Did FOI include baseline infections? Not clear from the definition in subsection “Data analysis”. Please state clearly at the first instance of the term. I assume baseline infections are not included in FOI since it is not possible to know when these started.

Subsection “Rate of clearance of infection and duration of infection by sex”: "68/105 (64.8%) baseline infections in males…"- as above, please state the proportion of all infections included in the analysis that were baseline infections, for males and females separately.

Discussion: "stronger effect in adults than in school-aged children, which was not seen in this cohort". Studies in adolescents have shown a strong relationship between control of parasitemia and adrenarche/DHEAS levels. The authors fail to cite the existing literature showing adrenarche and increasing DHEAS levels which correspond to malaria resistance in males and females. Citing these studies will contribute to the findings here, inasmuch as adrenarche starts earlier in females than males, and the relationship of DHEAS to malaria resistance carries over an extended age window.

Discussion: "a unique area with previously very high transmission intensity that has been greatly reduced in recent years by repeated rounds of IRS." This is not unique. The Garki project provides a similar example. In the Garki project report (p 155), there is no prominent difference in males vs females before the intervention; further, there is a prominent difference after the intervention in the intervention communities, but not in the control communities. This seems relevant to the current study and should be discussed. The Garki project is specifically called out in paragraph three of the Discussion.

Subsection “Data analysis”: "Infections were censored…" Censored for the clearance analysis only, or censored for both FOI and clearance analysis? Please state clearly.

Table 1: "Episodes of malaria**, (incidence*)" should just be "Episodes of malaria**", correct?

2) The authors point out that there is a low likelihood of reinfection with the same genotype. Addressing the issue of clonotype distribution within and between households may not be possible to include easily in this study. For that reason, please consider adding a note of this limitation to the discussion, since your analysis of the duration of infections (which relies on time infected with the same clonotype) may be impacted if there were a strong tendency of individuals to be reinfected with the same clones.

3) Please consider adding the following:

a) The literature appears conflicting, with significant confounding factors (such as treatment seeking) and study design. There may also be a species difference. See Tjitra et al. PlosMed 2008 Figure 2.

b) Inherent genetic factors remain a plausible explanation and X-linked disorders would warrant particular scrutiny. One point of correct for G6PD, is that although males are at higher risk of severe deficiency (hemizygotes ~5%), the greatest proportion are

actually heterozygous females with intermediate deficiency (15-20%). If intermediate deficiency is sufficient to effect clinical susceptibility to disease then this could be an interesting sub group.

eLife. 2020 Oct 27;9:e59872. doi: 10.7554/eLife.59872.sa2

Author response


Essential revisions:

A general question for this study is the scope and generalizability of the problem the authors are addressing; while many earlier studies reported malaria epidemiology, only a handful reported sexual dimorphism in risk. The papers cited by the authors describe different (and inconsistent) evidence for sexual dimorphism. Pathak et al., 2012 cited in the paper states: "Sexual dimorphism does not exist in hyperendemic regions for both Plasmodium falciparum and P. vivax infections, although some reports note an increased parasite density in pubertal and post-pubertal males".

We have expanded the Introduction to better review the evidence. Although sex-specific differences in malaria have not been uniformly described in younger children, the preponderance of the evidence over all regions is in favor of a male bias in malaria infections in school-aged children and adults (excepting pregnant women). The comprehensive Garki study (Molineaux et al., 1980) found that after 5 years of age, males have higher average parasite prevalence than females; several of the differences were statistically significant, including a greater than 2-fold higher prevalence of P. falciparum in the age-groups 9-18. Similarly, studies in Ghana (Landgraf et al., 1994), Sudan (Creasy 2004), and India (Pathak et al., 2012) have reported higher test positivity rates in school-aged children when comparing males with females. In addition, the majority of studies on malaria incidence and/or prevalence that stratify by sex in adulthood found a male bias in the observed measure of disease (Molineaux et al., 1980, Meek, 1988; Abdalla, Malik, and Ali, 2007; Pathak et al., 2012). Other studies have found a male bias in malaria incidence and/or prevalence but do not explicitly stratify by age group (Carmago 1996, Moon and ho, 2001; Mulu et al., 2013). Less commonly, researchers have found no sex-specific difference in adulthood in burden of malaria (Mendis et al., 1990, Sur 2006). Unfortunately, many papers do not stratify malaria prevalence or incidence by sex as they primarily focus on age stratification. In addition, the data that are published on malaria prevalence are often based on microscopy data alone. For example, in our data, Table 1 does not show a statistically significant difference between microscopy prevalence by sex using a simple comparison of proportions but does show a significant difference by qPCR prevalence, which is a much more sensitive diagnostic. If in fact higher malaria prevalence in males is driven by longer infections, then the tail end of these infections would most likely be lower density and more difficult to detect without molecular diagnostics.

Issues with the age categorization: why were these age bands chosen? If hormonal differences play a role in gender differences then surely this would increase after puberty. Perhaps it did and that is why higher difference is emerging in the 5-15 group. The three categories lose important data on the age sex distributions that could be highly relevant to the analysis. Surely 5-12yrs would have helped explore the puberty issue. Another approach would be to retain age as a continuous variable. Either way this doesn't explain the smaller difference in adults – although there were low numbers of males and few events. Did the age limit extend into older/menopausal women and might this have contributed to the findings? Please discuss these observations in the conclusions. The discussion also needs to acknowledge the limitations of the study and how assumptions and residual confounding may have impacted upon the findings.

Swimmers plots or similar figure may be helpful to display duration of individual infections over time. These could be stratified by sex, would show the number and durations of "infections" and of "clones", and individuals could be sorted by age so that the relevant age groups could be indicated.

First, we apologize that the last 2 paragraphs of the manuscript were somehow lost when formatting for submission (this included the limitations paragraph, which we have now expanded to two paragraphs based on reviewers’ comments).

Age groups were initially chosen to be consistent with numerous other publications we and others have published on this and similar cohorts, and to reflect salient differences in the overall epidemiology of disease between young children, school-age children, and adults. When changing the adjusted model so that age in continuous years is used instead of age categories, there is no meaningful change to the finding of faster clearance of infection in females (HR = 1.89 by clone and HR = 1.96 by infection event); however this does not answer the question of interaction by age. Based on the reviewers’ comments, we have also performed the analysis with age bands <8 years of age, 8-13 years, and 13+ years given that the onset of puberty occurs on average at approximately age 8-9 in females and age 9-10 in males in the US and may occur slightly later in Uganda. Changing the age bands did not meaningfully alter our findings (adjusted hazard ratio (HR) for clearance of infection by clone in females vs males= 1.64, adjusted HR by infection event = 2.13). To look for evidence of interaction by age, we stratified by these modified age bands and calculated unadjusted HR for clearance of infection in females versus males. By clone, unadjusted hazard ratio (HR) for clearance of infection in females versus males was 1.69 (0.45 – 6.34) in children <8, 2.01 (0.81 – 5.00) in children 8-13, and 1.58 (0.95 – 2.63) in those over the age of 13. Performing the same unadjusted stratified analysis for the age bands used in our paper gave HR for clearance of clones in females vs males of 2.31 (0.73-7.28) in children <5, 1.92 (0.97 – 3.78) in children 5-15, and 1.42 (0.70 – 2.87) in those 16 and older. By infection event, HR for clearance of infection using modified age bands was 0.79 (0.23 – 2.73) in children <8, 1.56 (0.19 – 12.49) in children 8-13, and 3.54 (1.68 – 7.48) in those over 13. Performing the same unadjusted stratified analysis for the age bands used in our paper, HR for clearance of infection events in females was 1.09 (0.31-3.78) in children < 5, 1.50 (0.42 – 5.36) in children 5-15, and 3.79 (1.48. – 9.69) in those 16 and older. Regardless of whether the original age bands or the modified age bands are used, similar trends are seen. The trend toward more evidence of sexual dimorphism in the highest age group by infection event but not by clone is likely explained by the fact that adults are known to have infections with significantly lower parasite densities than children; while we can detect infections down to 0.05 parasites/microliter using ultrasensitive qPCR, we are only able to genotype infections at >=0.1 parasite/microliter. While there is no definitive evidence for an interaction between age and sex when it is included in the final adjusted model, our sample size limits the power that we have to detect this type of interaction. Finally, as pointed out by the reviewers, post-menopausal women were included in the highest age band which could confound the findings in that age category, but are not of sufficient numbers to make any meaningful conclusions in subgroup analysis. We added these points to the Discussion.

Please include information if baseline infections differed by sex. Baseline infections have significantly longer duration than new infections (which is counter-intuitive to this reviewer) and constituted a majority of the infections in their analyses. Did the authors see similar patterns of sexual dimorphism in their baseline infections as well as their new infections?

Information on baseline infections by sex has now been added to the manuscript in two places in the Results. Males had more baseline infections by clone but not by infection event, which indicates higher complexity of infection in their baseline infections, consistent with our finding of longer duration of infection.

Data on important confounders are missing, for example, genetics/ hemoglobinopathies, drug use (by report and by drug levels during "asymptomatic" infections), time since residual spraying if done.

We do not have data available on genetic hemoglobinopathies that affect susceptibility to malaria for this cohort. G6PD deficiency, which is an X-linked recessive disorder, is more common in males. There is no reason to suspect a sex bias in sickle-cell anemia (or trait) or α thalassemia since they are not sex-linked traits; although this could be an empirical confounder if there was an unequal distribution by random chance in this cohort. We have added this to the study limitations as mentioned in the Discussion. Likewise, we do not have data on drug levels in this cohort. However, our clinic is open 7 days a week, free, high-quality care is provided, and transport is reimbursed, giving participants a little incentive for seeking outside care. We have performed studies of this design for many years and receipt of outside care has never been a major issue. We do ask all participants at every routine visit if they have taken any outside antimalarials. Only for 4 participants was this ever reported (only once per participant). I have added a line acknowledging this in the Results. All IRS in this district was performed over 1 month and so there was little difference between households in time since residual spraying.

Terms used in the paper have overlapping meanings ("clone", "infection", "parasitemia", "malaria infection", "malaria", "asymptomatic malaria", "symptomatic malaria", etc) and not all are clearly defined-would be good to define each and use consistently. The definitions for "clearance" require some clarification.

We apologize for any confusion in our terminology. An infection with malaria parasites (measured as detectable parasitemia) can be caused by a single strain (which we refer to as clone) or by multiple clones. Analyses can be performed at the clone level or at the infection-event level (which groups various clones into a single infection); therefore it is necessary to use the word “infection” at a more general level at times. I have made clarifying edits regarding clone/infection throughout the paper in the Materials and methods and Results. “Asymptomatic malaria” is not a term used in our paper. Episodes of clinical malaria are clearly defined in the Materials and methods and in Table 1. Parasitemia is used to denote the presence of malaria parasites in the blood and is used only in the context of prevalence of infection (Table 1 and first paragraph of Results).

Other major concerns surround the robustness of the conclusions based on the statistical analysis:

1) No direct statistical test is given to show that prevalence differs between the sexes after adjusting for behavioural factors (and age etc).

Using generalized estimating equations to account for clustering by individual, prevalence ratio (PR) of P. falciparum parasitemia by microscopy in females versus males across all age categories was 0.49 (95% CI 0.26-0.90, p = 0.02), with relative differences in prevalence most pronounced in the oldest age group. Similar findings were seen when prevalence was assessed by ultrasensitive qPCR, with PR = 0.64 in females vs. males (95% CI 0.43-0.96, p = 0.03), again with the largest differences seen in the oldest age group. Including age category, LLIN use, and travel in a Poisson regression model using generalized estimating equations to account for clustering by individual did not qualitatively change the parasite prevalence ratio in females versus males, for microscopic parasitemia (PR in females vs. males = 0.57, 95% CI 0.42-0.77, p <0.001) or for submicroscopic parasitemia (PR in females vs. males = 0.67, 95% CI 0.60-0.76, p <0.001). This is addressed now in the text. In fact, including behavioral factors in the model increased QIC (decreased model fit), with the majority of variance in the model explained by age category and sex alone.

2) The analysis of infection clearance and duration of infection (and even the FOI) are all based on the arbitrary assumption that 4 negative routine visits are needed before an infection is deemed to have been cleared. How many of the results change when this assumption is relaxed?

We thank the reviewers for this raising this important point, as this is something we have spent a lot of time discussing and trying to figure out. While unfortunately the sample size of this study (and the number of observed infections) limits our capacity to model the clearance and re-infection processes jointly, the assumption allowing a maximum of 3 skips to be considered the same infection was not entirely arbitrary, as now discussed in the text. The decision was motivated by competing probabilities of having not detected a persistent infection at certain timepoints (low detectability) versus a new infection occurring, considering that: 1) diversity of the genotyped locus was quite high, resulting in a low probability (5.1%) for the different parasites having the same genotype by chance; 2) force of infection was low in the study, making it less likely that infection by an identical clone was new vs persistent; and 3) as a result parasite density was low in these long duration, “old” infections, making it more probable that persistent infections present would drop below the limit of detection and be missed at multiple timepoints. The reviewer’s suggestion to perform a sensitivity analysis with a fewer number of skips is a good suggestion. We have now performed this, and the main results do not change (i.e., there is still a faster clearance of infection in females) when the data are analyzed allowing for 2 skips or 1 skip. These data are now referred to in the manuscript and available in the Supplementary file 1. Notably, multiple prior longitudinal genotyping studies have made the same assumption that re-infection with the same clone is relatively rare, which is reasonable given the high genetic diversity seen in this setting and low force of infection; these are referred to in the Discussion.

3) The analysis of FOI between sexes may be confounded by what is classed as a "new infection". Does this result still hold when the definition of new infection is altered?

Fundamentally, when you change the number of skips allowed, some of those clones that were previously considered part of a long infection will become classified as new infections (likely falsely, given the considerations above). Thus, FOI increases for all groups. Tables showing FOI for 2 skips and 1 skip are provided in Supplementary file 1. There remains a trend toward higher FOI in males when Table 2 is replicated with 2 skips or 1 skip, but even when performing the analysis with 1 skip, there is no significant difference in FOI between the sexes (IRR for females vs. males when adjusted for age category = 0.71, 95% CI 0.41 to 1.24 by clone and IRR = 0.75, 95% CI 0.47 to 1.22 by infection event). There may be a difference by sex in the force of infection that we are unable to detect due to small number of new infections (overall low FOI). Notably, no matter how many skips are chosen, there remains sexual dimorphism in duration of infection.

Also, we would like to note a small correction to Table 2; we had mistakenly not accounted for repeated measures in our original estimation of FOI. The results have changed very slightly on account of this, and the Materials and methods have been updated to reflect the correct technique used (Poisson regression with generalized estimating equations).

These are suggestions for validating the robustness of the conclusions.

1) The authors do not directly test whether sex is still a predictor of higher prevalence after other factors are considered, which one would expect to see in the typical analysis in order to make the claim that behavioural factors do not explain the difference between men and women. Instead, the authors argue by looking at a table of the mean characteristics of the population what is and is not different between men and women (Table 1). Their conclusion on the basis of a lack of significance between behavioural pattern beteween sexes this is not a statistical argument. If the authors wish to show that, after accounting for behavioural differences, there is still a link between sex and prevalence, they need a full statistical model that includes all factors. The authors should run a logistic regression looking at the predictors of infection status in individuals, and showing that sex is a significant predictor. Covariates in this model should be sex, age, the behavioural data (e.g. bednet usage, travel, etc), time point. The model should also include random-effects to account for repeated measures within individuals. If no other factors are included, it could be assumed that sex would be a predictor of infection. However, forwards and/or backwards regression could then be used to ask the question "is sex a significant predictor of infection once adjusting for other behavioural factors are included in the model?" Similar analysis is performed on the clearance data so this should be easily achievable for the authors. If sex is not a predictor of infection after adjusting for age and behavioural factors, then one might conclude that risk of detecting an infection is well explained by behavioural factors and not sex differences. If sex is still required to explain the data then that would be evidence for a difference in prevalence that is sex based on not behavioural.

This is addressed above under response #1.

2) The authors argue that women have faster clearance times of asymptotic infection but this analysis is directly dependent on the definition of when an infection is cleared. Follow-up is every 28 days in this study and the end of an infection was defined as a clonotype being absent for 4 routine visits (i.e. someone might have a clone present at one visit, and at a visit 112 days later, but not detected at all in-between and it would be considered the same infection). If one is reinfected with the same clone (which might be very probably in an environment where 10 of the clones detected account for 55% of all observed clones) it is possible to count a new infection as one long infection. This issue may both artificially prolonging clearance times in the analysis and reduce the number of "new infections" (i.e. lower the FOI calculated) and may bias observations for men and women. This demand for 4 consecutive negative visits before an infection is deemed to have been cleared is an arbitrary choice and to be believable the conclusions of the analysis should not depend strongly on this assumption. For this reason, the authors must check that varying this criteria doesn't alter the results too strongly.

This has been addressed above.

a) The authors should explore how this definition of clearance (being 4 negative routine visits) impacts the conclusions. What if only one negative time point was used as the definition of the end of an infection? How would this affect the FOI for men and women, and how would it impact the clearance times observed? Since it is not possible to really know when an infection ended, and a new infection began, it is important to test that the conclusions are robust to these assumptions.

These points are addressed above.

b) It may be that behaviour patterns impact the likelihood of being infected with the same clonotype twice. For example, if one group does more travel than another group, then they may be more likely to be infected with different clonotypes. This diversity of clonotypes would make an individual look like they had shorter durations of infection compared with a group that travels less and is re-infected over and over with the same clonotype. i.e. Women who travelled more may only look like they have shorter infection because they are not re-infected with the same clone as much as the men who travelled less. To test this, the authors should look at the distribution of clonotype frequencies and diversity between men and women (e.g. do women tend to have infection with a greater diversity in clonotypes than men?). This might also make men look like they have a lower FOI than they really do, and women a higher FOI than they really do, and it may make it seem like the two groups have the same FOI when actually they are very different.

The reviewer raises an important point. However, there was actually a slightly greater diversity of clones in men, with 34 unique clones found in men and 26 in women. Furthermore, only women over the age of 15 traveled more than males, so this is unlikely to distort the overall findings as proposed.

c) The issue of 4 negatives being required for "clearance" may explain why the infection present in individuals at enrollment is longer than "new infection". This result would change if the end of an infection was defined as the first negative rather than requiring 4 negatives. This may indicate a bias in the analysis towards counting things as one infection that are actually reinfection with the same clonotype.

This was addressed above.

3) The authors should confirm that there is no difference in FOI between the sexes. Even though not significant, there appears to be a trend towards higher FOI in men compared with women in all three age groups (Table 2). This is important because it is not clear how different the FOI in men and women would need to be in order for it to manifest in the differences in prevalence that exist between men and women. It may be that there is really a difference in prevalence observed and that this study does not have statistical power to detect such a difference. Also, calculating FOI as the authors have done requires knowing "new infections" as opposed to just the carriage of old infection, and the definition of what is a new infection is going to influence these conclusions. The authors should further validate this conclusion from their data by: (a) changing their definition of new infection (as in point 2) and recalculating the FOI and determining if there is any significant difference between the sexes. (b) performing a time-to-next-infection analysis (survival analysis, e.g. cox regression), rather than calculate the FOI as new infections over person time. The latter would involve looking at the time until the next new clone appears in each person in a survival model (with appropriate censoring). The question would then be "is sex a significant covariate in a survival model of time to next infection (after including other behavioural factors as well, such as travel)?". If these two other approaches also reveal no difference in FOI between the sexes, then the conclusion that there is no difference in infection pressure on the sexes would be more robust/convincing.

We have performed a sensitivity analysis as discussed in Response #8. Performing a time-to-next-infection analysis as suggested by the reviewer, adjusting for sex and age category and clustering for random effects by participant, results in a HR of new infections in males to females of 1.26 (0.64 – 2.46) and 1.12 (0.76-1.65) by clone. As above, though there is a trend toward more new infections in men, there is no statistical power to detect a difference in this cohort. Since these results are merely another way of analyzing the same data as in the current manuscript and show the same result, these results were not added to the manuscript.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The information on baseline infections, a potential confounder to these analyses, still needs further clarification.

Subsection “Cohort participants and P. falciparum infections”: "25 participants.… were enrolled after initial enrollment…" Please state whether their baseline data contributed to analyses of baseline infections. It seems so from subsection “Data analysis” but should be stated explicitly.

Yes, their baseline data also contributed to the analyses of baseline infections because any infection detected in the first 60 days after their enrollment (whenever that enrollment was) would still be a baseline infection. This is now clarified in the manuscript as follows:

“Data from the 25 dynamically enrolled participants contributed to all analyses, including that of baseline infections.”

Subsection “Cohort participants and P. falciparum infections” paragraph two: "35 samples had very low density infections (< 1 parasite/μL) that could not be genotyped and had infections characterized at the event level only". Did these come from 35 unique individuals who had a single infection? And do these represent the difference between 149 individuals with infections versus 114 individuals who had 822 typable samples? If so, please state this explicitly. It appears so from Figure 1.

Yes, they did come from 35 unique individuals. This has been clarified in the manuscript as follows:

“114 participants had 822 successfully genotyped samples and had infections characterized by clone and by infection event. 35 samples (from 35 unique participants) had very low-density infections (< 1 parasite/µL) that could not be genotyped; these infections were characterized at the infection event level only. We achieved a read count of >10,000 for 92% of genotyped samples, identifying 45 unique AMA-1 clones in our population (frequencies and sequences in Supplementary file 1E).”

"117/185.… baseline infections in males". Please state the proportion of all infections included in the analysis that were baseline infections, for males and females separately, at the clone and the infection event level. Baseline infections are an important confounder in this study because they differ from other infections and disproportionately occur in males at the clone level.

I have changed the denominators so that the proportion of baseline infections out of all infections in males vs females is now stated as follows:

“At the clone level, the proportion of baseline infections out of all infections in males was 117/171 (68.4%), compared to 68/116 (58.8%) baseline infections in females (p = 0.10). At the infection event level, there were 54/104 (51.9%) baseline infections in males and 45/89 (50.6%) baseline infections in females (p = 0.89).”

Subsection “Behavioral malaria risk factors and measures of malaria burden”: "differences in prevalence…" Prevalence is defined in subsection “Data analysis”, but it might be helpful to use the term "period prevalence" or a simple explanation here, since many non-epidemiology readers may assume that prevalence here refers to a point prevalence.

I have added a line of explanation as follows:

“Microscopic parasite prevalence was defined by the number of smear-positive routine visits over all routine visits. Parasite prevalence by qPCR was defined as the number of qPCR-positive visits over all routine visits. Thus, these measures represent the average prevalence during routine visits.”

This is not period prevalence since the denominator is not number of individuals; it is instead the average parasite prevalence during routine visits.

In the same section please define COI.

This has now been defined at its earliest appearance.

Subsection “Force of infection by age and sex”: "Force of infection…" Did FOI include baseline infections? Not clear from the definition in subsection “Data analysis”. Please state clearly at the first instance of the term. I assume baseline infections are not included in FOI since it is not possible to know when these started.

By definition, FOI is a measure wherein the numerator is only new infections. Therefore baseline infections are not included in FOI, as new infections are defined as infections detected in a participant after day 60. I have attempted to clarify further in the Materials and methods in several places and added the additional line:

“Force of infection (FOI) was defined as the number of new infections, including malaria episodes, divided by person time. Since the start of baseline infections was not observed, baseline infections were not included in calculating FOI.”

Subsection “Rate of clearance of infection and duration of infection by sex”: "68/105 (64.8%) baseline infections in males…"- as above, please state the proportion of all infections included in the analysis that were baseline infections, for males and females separately.

This is now corrected, as follows:

“At the clone level, 105 baseline infections and 53 new infections were included; there was a slightly higher proportion of baseline infections in males (68/99, 68.7%), compared to the proportion of baseline infections in females (37/59, 62.7%) (p = 0.49). At the infection event level, 58 baseline infections and 51 new infections were included and there was no difference in the proportion of baseline infections by sex, with 32/60 (53.3%) baseline infections in males and 26/49 (53.1%) baseline infections in females (p = 1.0).”

Discussion: "stronger effect in adults than in school-aged children, which was not seen in this cohort". Studies in adolescents have shown a strong relationship between control of parasitemia and adrenarche/DHEAS levels. The authors fail to cite the existing literature showing adrenarche and increasing DHEAS levels which correspond to malaria resistance in males and females. Citing these studies will contribute to the findings here, inasmuch as adrenarche starts earlier in females than males, and the relationship of DHEAS to malaria resistance carries over an extended age window.

I have cited this literature now in the Discussion with its relevance to our study.

Discussion: "a unique area with previously very high transmission intensity that has been greatly reduced in recent years by repeated rounds of IRS." This is not unique. The Garki project provides a similar example. In the Garki project report (p 155), there is no prominent difference in males vs females before the intervention; further, there is a prominent difference after the intervention in the intervention communities, but not in the control communities. This seems relevant to the current study and should be discussed. The Garki project is specifically called out in paragraph three of the Discussion.

By the word “unique” I had meant that it is a specific type of setting, one in which transmission was previously very high and is now very low. I have changed the word “unique” to “specific setting”. Thank you for pointing out additional findings in the Garki study. I had missed this subtlety regarding widening of the prevalence gap between males and females after control measures, and have now incorporated them in the Discussion as follows. It is very interesting that we may be seeing a similar phenomenon.

“Very few studies have been conducted to explore immunological differences between males and females in their response to the malaria parasite. […] More studies are needed to elucidate the relationship between sex-based biological differences between males and females and their impact on the development of effective antimalarial immunity in humans.”

Subsection “Data analysis”: "Infections were censored…" Censored for the clearance analysis only, or censored for both FOI and clearance analysis? Please state clearly.

Censored for the clearance analysis only, as now clarified in the Materials and methods as follows:

“Infections were censored for this analysis if they were only observed in the first three months or the last three months (before January 01, 2018 and after January 01, 2019) because they were not observed for long enough to determine whether clearance occurred.”

Table 1: "Episodes of malaria**, (incidence*)" should just be "Episodes of malaria**", correct?

This has been corrected, thank you for noticing.

2) The authors point out that there is a low likelihood of reinfection with the same genotype. Addressing the issue of clonotype distribution within and between households may not be possible to include easily in this study. For that reason, please consider adding a note of this limitation to the Discussion, since your analysis of the duration of infections (which relies on time infected with the same clonotype) may be impacted if there were a strong tendency of individuals to be reinfected with the same clones.

This is now addressed as follows:

“There were not enough infections to perform a rigorous analysis of the distribution of clones within and between households, but given that the overall force of infection was quite low, the probability of re-infection with the same clone already present in a participant from another member in the household (which could bias toward longer duration of infection) was low and unlikely to have introduced any significant bias by sex.”

3) Please consider adding the following:

a) The literature appears conflicting, with significant confounding factors (such as treatment seeking) and study design. There may also be a species difference. See Tjitra et al. PlosMed 2008 Figure 2.

We have added an acknowledgment that some studies do not show a male bias in prevalence or incidence of malaria in the Discussion. We did not choose to add the Tjitra et al., 2008 reference, as Figure 2 is specifically referring to severe malaria and we are not addressing the severity of malaria in females versus males in this paper, only uncomplicated malaria.

“Though there are some conflicting reports in the literature, the majority of studies of malaria incidence and/or prevalence that evaluated associations with sex in late childhood, adolescence and adulthood have found a male bias in the observed measure of burden10–16. We note that this is more often observed in hypoendemic settings and may be confounded by factors such as treatment-seeking behavior; however, this male bias has been reported in studies of both P. vivax and P. falciparum. Overall, these studies consistently suggest that males exhibit higher incidence and/or prevalence of malaria that begins during late childhood, persisting through puberty and the majority of adulthood (excepting the years when pregnancy puts women at higher risk).”

b) Inherent genetic factors remain a plausible explanation and X-linked disorders would warrant particular scrutiny. One point of correct for G6PD, is that although males are at higher risk of severe deficiency (hemizygotes ~5%), the greatest proportion are

actually heterozygous females with intermediate deficiency (15-20%). If intermediate deficiency is sufficient to effect clinical susceptibility to disease then this could be an interesting sub group.

Genetic hemoglobinopathies (which would include G6PD) are an unmeasured confounder acknowledged in the Discussion. We realize this is a limitation and will plan to generate data on genetic hemoglobinopathies in future analyses looking at sex-based differences in malaria infection.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Dorsey G. 2020. PRISM2 ICEMR Cohort. ClinEpiDB. DS_51b40fe2e2

    Supplementary Materials

    Supplementary file 1. (a) AMA-1 hemi-nested PCR protocol for amplicon deep-sequencing. (b) Bioinformatics workflow. (c) Declining qPCR density over time in the cohort. (d) Detailed explanation of skip rule criteria. (e) Haplotype sequences and frequencies. (f) Sensitivity analysis of molecular force of infection: Table 2 replicated using 2 skips or 1 skip. (g) Sensitivity analysis of duration of infection: Table 3 replicated using 2 skips or 1 skip.
    elife-59872-supp1.docx (293.7KB, docx)
    Transparent reporting form

    Data Availability Statement

    Data from the cohort study is available through an open-access clinical epidemiology database resource, ClinEpiDB at https://clinepidb.org/ce/app/record/dataset/DS_51b40fe2e2. Genotyping data and code used to generate tables and figures is available on GitHub (Briggs, 2020; copy archived at swh:1:rev:cf6c3256e609f4f136fc8d90f9cae1d61d6d8d63).

    Data from the PRISM2 cohort study is available through a novel open-access clinical epidemiology database resource here: https://clinepidb.org/ce/app/record/dataset/DS_51b40fe2e2. Sequencing data is available on Github at https://github.com/EPPIcenter/sex_based_differences as referenced in the paper. In addition. all sequences of haplotypes are included in Supplementary file 1.

    The following dataset was generated:

    Dorsey G. 2020. PRISM2 ICEMR Cohort. ClinEpiDB. DS_51b40fe2e2


    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES