Ethics statement
This study was approved by the Brazilian National Committee of Ethics (CONEP) (349.211/2013) and the Committee of Ethics for Clinical Investigation of the Barcelona Hospital Clinic (7306/2012). All participants were informed about the objectives of the study as well as the potential risks and benefits of their participation in the study. An informed consent form was signed by all study participants or by a parent or legal guardian in case of participants that were under 18 years of age. Children between 12 and 17 years of age signed an additional assent form. As routinely done, malaria patients received treatment with 25 mg/kg of chloroquine over a 3-day period (10 mg/kg on day 0 and 7.5 mg/kg on days 1 and 2. Primaquine was prescribed at the dosage of 0.5 mg/kg per day, during 7 days.
Study design and subjects
This cohort study was conducted in the Brasileirinho, Ipiranga and Puraquequara communities, located in Manaus peri-urban area, between April 2013 and March 2014 (Figure 1A). A detailed description of the study area has been presented elsewhere [17]. According to a census performed by a Fundação de Medicina Tropical Doutor Heitor Vieira Dourado (FMT-HVD) field team, the population of the study area was estimated to be approximately 2,400 inhabitants before the start of the study, in 2012. Each community has access to a malaria clinic for microscopy-based malaria diagnosis and treatment. A total of 1,200 participants of any age were enrolled into the study in April 2013.
Data and sample collection
For each study participant, a questionnaire was completed containing personal information such as age, gender, occupation, pregnancy, history of travel, as well as information on malaria preventive measures, previous malaria episodes and current health status. Upon enrolment and monthly during follow-up, finger-prick blood 151 samples (~300 μL) were collected using Microtainer® tubes containing EDTA and sodium fluoride (Becton Dickinson, USA). In infants, blood was obtained by either heel or toe puncture. Within one hour of collection, 50 μL of blood were transferred into a reaction tube containing 250 μL of RNAprotect (QIAGEN, Germany) in order to preserve RNA for downstream analyses [22] and 200 μL of whole blood was transferred to another reaction tube. Samples were stored in cooling boxes until arrival in the laboratory, where the 200 μL sample was separated into a red blood cell (RBC) pellet and plasma. All samples were frozen at -80ºC until further processing. If the collected blood volume was <250 μL, the actual volume was recorded.
Clinical symptoms
In the case of symptoms related to malaria or in 161 the case of increased body temperature (>37.5ºC), a thick blood smear (TBS) was prepared according to World Health Organization guidelines [23]. When positive for malaria, appropriate treatment was provided in accordance with the national guidelines of the Brazilian Ministry of Health [24]. An asymptomatic infection was defined as presence of a malarial infection by TBS, but absence of fever and any other malaria related symptoms (chills, sweating, headache, vomit, abdominal pain) at the moment of sample collection, or anytime in the preceding 48 hours.
Plasmodium spp. infection, clones and gametocyte carriage
Pelleted RBCs, obtained from 200 μL of whole blood, were re-suspended in PBS and genomic DNA was extracted using FavorPrep 96-well Genomic DNA Kit (Favorgen, Taiwan) according to the manufacturer’s instructions. DNA was eluted with 2x 100 μL of elution buffer and stored at -20ºC until assayed by PCR. If the amount of whole blood available for DNA extraction was ≤100 μL, the DNA volume was vacuum concentrated until reaching the original blood volume recorded. RNA from 50 μL whole blood, stored in RNAprotect, was extracted using RNeasy Plus 96-well kit (QIAGEN, Germany) and eluted in 50 μL RNase-free dH2O as described previously [22]. All DNA samples were subject to a generic Plasmodium species (QMAL) qPCR targeting a conserved region of the 18S rRNA gene [22]. QMAL-positive samples were further analysed by species-specific qPCR assays targeting the18S rRNA genes of P. falciparum and P. vivax, as previously described [22,25]. For detection of P. falciparum, a modified reverse primer was used [26]. For quantification of 18S rRNA gene copy numbers, in each experiment three dilutions of control plasmids containing the respective amplicons were included in triplicates (102, 104 and 106 copies/μL). For genotyping individual P. vivax clones, the molecular markers msp1F3 and MS16 were typed using capillary electrophoresis for highly precise fragment sizing allowing for longitudinal follow up of individual parasite clones. Details of the genotyping methods have been described previously [27]. RT-qPCR assays were performed on RNAs from all P. vivax and/or P. falciparum positive samples to detect gametocyte-specific transcripts of the pvs25 (P. vivax) and pfs25 (P. falciparum) genes. For quantification of pvs25 and pfs25 transcript numbers, control plasmids containing the amplified region were included as standards in each run. All qPCR and RT-qPCR assays were performed on a 7500 Fast Real-Time PCR System (Applied Biosystems).
Statistical analysis
Data from questionnaires were imported into databases using Cardiff TeleForm version 10.4.1 (Cardiff Software). Individual databases were combined in Microsoft Access 2010. For incidence calculations (molFOB and clinical malaria incidence), subject data were censored on the last visit before two consecutively missed scheduled follow-up visits in order to reduce bias [28]. Differences in proportions were tested for statistical significance using the McNemar X2 test with continuity correction. To achieve normal distribution, qPCR densities were expressed as log10-transformed 18S rRNA genomic copies/μL blood for asexual parasites, and log10-transformed pfs25 or pvs25 transcripts/μL blood for gametocytes. Geometric means of densities were calculated. Differences in densities of asexual or sexual-stage parasites were tested for statistical significance using Welch’s two-sample t-test.
The molecular force of new P. vivax blood-stage infections (molFOB) was calculated by counting the number of genotypes observed at each visit, that had not been present in the preceding two visits (0-0-1 patterns) and adjusting these counts by the respective times-at-risk. molFOB for P. vivax was determined for both genetic markers combined. Negative binomial regression models were used to assess the influence of different risk-factors on the incidence of P. vivax and P. falciparum gametocyte positivity as previously described, using positivity counts and times-at-risk over the entire period of observation [28]. Since molFOB is a count variable measured per individual over a specific exposure time (time at risk), and is overdispersed, a negative binomial regression model was chosen in which the exposure time at risk is used as offset. If we define μj as the log of the number of genotypes at visit j, then for each infection pattern j (0-0-0 or 0-0-1) we have μj = exp(βxj + offsetj + νj), where β is a vector of regression coefficients, offsetj=log(exposure time) and νj follows a gamma distribution (to give a negative binomial distribution). Incidence rate ratios (IRR) and adjusted IRR (aIRR) were calculated with their respective 95% confidence intervals. Because using the collapsed data to model molFOB for each individual does not allow for the analysis of time-changing covariates, factors influencing frequency of parasite positivity and frequency of clinical episodes within the study period were explored using multiple failure time models allowing for time-changing covariates [29]. For multiple failure time models, hazard rate ratios (HRR) are calculated with 95% confidence intervals. In these models, parasite positivity and clinical episodes were equivalent to a ‘failed’ outcome, respectively. In addition to the adjusted statistical models presented in the main manuscript, univariate analyses and multivariate analyses with backward selection are provided as Supplemental Materials. Versions of each model analysis were implemented with backwards selection to eliminate non-significant covariates resulting in the most parsimonious models. Statistical analyses were conducted using R v3.1.1 or STATA v14.