SEARCH

SEARCH BY CITATION

Keywords:

  • Malaria;
  • Plasmodium falciparum ;
  • Disease Reservoirs;
  • Risk;
  • Algorithms;
  • Epidemiology;
  • India

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Objective

To characterise the epidemiology of Plasmodium falciparum gametocytemia and determine the prevalence, age structure and the viability of a predictive model for detection.

Methods

We collected data from 21 therapeutic efficacy trials conducted in India during 2009–2010 and estimated the contribution of each age group to the reservoir of transmission. We built a predictive model for gametocytemia and calculated the diagnostic utility of different score cut-offs from our risk score.

Results

Gametocytemia was present in 18% (248/1 335) of patients and decreased with age. Adults constituted 43%, school-age children 45% and under fives 12% of the reservoir for potential transmission. Our model retained age, sex, region and previous antimalarial drug intake as predictors of gametocytemia. The area under the receiver operator characteristic curve was 0.76 (95%CI:0.73,0.78), and a cut-off of 14 or more on a risk score ranging from 0 to 46 provided 91% (95%CI:88,95) sensitivity and 33% (95%CI:31,36) specificity for detecting gametocytemia.

Conclusions

Gametocytemia was common in India and varied by region. Notably, adults contributed substantially to the reservoir for potential transmission. Predictive modelling to generate a clinical algorithm for detecting gametocytemia did not provide sufficient discrimination for targeting interventions.

Objectif

Caractériser l’épidémiologie de la gamétocytémie de Plasmodium falciparum et déterminer la prévalence, la structure d’âge et la viabilité d'un modèle prédictif pour la détection.

Méthodes

Nous avons recueilli des données provenant de 21 essais d'efficacité thérapeutiques réalisés en Inde durant la période 2009–2010 et avons estimé la contribution de chaque groupe d’âge au réservoir de la transmission. Nous avons construit un modèle prédictif pour la gamétocytémie et avons calculé l'utilité diagnostic de différents seuils de notre score de risque.

Résultats

La gamétocytémie était présente chez 18% (248/1.335) des patients et diminuait avec l’âge. Les adultes constituaient 43%, les enfants d’âge scolaire 45% et ceux de moins de cinq ans, 12% du réservoir pour la transmission potentielle. Notre modèle a retenu l’âge, le sexe, la région et la prise précédente d'un médicament antipaludique, comme facteurs prédictifs de la gamétocytémie. L'aire sous la courbe d'efficacité du récepteur opérateur était de 0,76 (IC95%: 0,73–0,78) et un seuil de 14 ou plus sur un score de risque allant de 0 à 46, procurait une sensibilité de 91% (IC95%: 88–95) et une spécificité de 33% (IC95%: 31–36) pour la détection de la gamétocytémie.

Conclusions

La gamétocytémie était courante en Inde et variait selon les régions. Notamment, les adultes contribuaient de façon substantielle au réservoir pour une potentielle transmission. La modélisation prédictive pour générer un algorithme clinique pour la détection de la gamétocytémie ne fournit pas une discrimination suffisante pour les interventions cibles.

Objetivo

Caracterizar la epidemiología de la gametocitemia de Plasmodium falciparum y determinar la prevalencia, la estructura de edad y la viabilidad de un modelo predictivo para la detección.

Métodos

Hemos recolectado datos de 21 ensayos de eficacia terapéutica realizados en India durante 2009–2010, y calculado la contribución de cada grupo de edad al reservorio de transmisión. Hemos construido un modelo predictivo para la gametocitemia y calculado la utilidad en el diagnóstico de diferentes puntos de corte de nuestra puntuación del riesgo.

Resultados

La gametocitemia estaba presente en un 18% (248/1,335) de los pacientes y disminuía con la edad. Los adultos constituían un 43%, los niños en edad escolar un 45% y los menores de doce años un 12% del reservorio para una transmisión potencial. Nuestro modelo retenía la edad, el sexo, la región y la toma previa de antimaláricos como vaticinadores de gametocitemia. El área bajo la curva ROC (Característica Operativa del Receptor) era de 0.76 (95%CI:0.73,0.78) y el punto de corte de 14 o más en la puntuación del riesgo con un rango entre 0 a 46 tenía un 91% (IC 95%:88,95) de sensibilidad y un 33% (IC95%:31,36) de especificidad para detectar la gametocitemia.

Conclusiones

La gametocitemia era común en la India y variaba según la región. De forma notable, los adultos contribuían sustancialmente al reservorio para una transmisión potencial. Los modelos predictivos para generar un algoritmo clínico para detectar la gametocitemia no proveían suficiente discriminación para aplicar en intervenciones.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Gametocytes are the sexual stage of Plasmodia that render the human host infectious to mosquitoes. Although gametocytes are central to understanding transmission, few studies have tried to demarcate the infectious reservoir (Bousema & Drakeley 2011). This may be because the infectious reservoir is not an important determinant of the intensity of transmission relative to high vectorial capacity in high transmission areas (Macdonald 1956). However, in low and moderate transmission areas, the proportion of infectious hosts is critical to the maintenance of endemicity (Macdonald 1956). Interventions for detecting and treating gametocytemia also differ from those used for asexual parasitaemia. So an improved understanding of the epidemiology of gametocytemia opens the possibility of distinct transmission-blocking control strategies.

Gametocyte density is generally lower than that of asexual parasites (typically less than 5% of the total parasite population), and levels below the detection limit of microscopy can infect mosquitoes (Taylor & Read 1997). The low density of gametocytes coupled with inadequate laboratory standards and the heavy workloads of technicians make the detection of gametocytemia difficult in routine settings. Recently, molecular methods that detect gametocyte stage-specific RNA transcripts have been employed in research (Schneider et al. 2004; Mlambo et al. 2008; Karl et al. 2009), but these may not be feasible for routine public health use. Associations of gametocytemia with easily discerned factors such as age, season and symptoms such as fever at the time of presentation could provide an alternative strategy for targeting gametocytocidal interventions. Clinical algorithms for predicting gametocytemia among diagnosed malaria patients could help improve its detection.

The control of malaria is a major challenge for India, which reported 1.1 million cases in 2012 (NVBDCP, National Vector Borne Disease Control Programme). The reduction in transmission is a priority; most of the countries have low or moderate malaria endemicity. The transmission of malaria in India is arguably the most complex in the world given the large geographic area, the presence of both major parasite species, a wide range of ecotypes and vectors and the enormous population (Kumar et al. 2007). In addition, lower acquired immunity, more adult malaria, better access to drugs, and mixed species infections alter the epidemiology of gametocytemia in India compared with sub-Saharan Africa (Singh et al. 2009). Sinton and others studied crescents primarily with respect to treatment and spleen size in the pre-independence era (Sinton 1926). Since 1990, only 4 published studies have described gametocytemia in India, but these had small sample sizes and limited coverage (Rajagopalan et al. 1990; Mohapatra et al. 1992, 1998; Kar et al. 2009). Most importantly, no study characterised the subpopulation with gametocytemia.

The goal of this study was to describe the epidemiology of P. falciparum gametocytemia in India and determine whether a clinical predictive model could improve its detection.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Study sites and population

We used data from 22 P. falciparum therapeutic efficacy trials conducted through the National Antimalarial Drug Resistance Monitoring Network of India in 2009 and 2010 (Figure 1) (Mishra et al. 2012). The National Vector Borne Disease Control Programme and the National Institute of Malaria Research purposively selected the study sites to represent P. falciparum transmission settings across the country. We could not obtain data from one site (Gadchiroli) because gametocytemia was not recorded in case record forms and slides could not be re-examined. We included all patients eligible for the World Health Organization (WHO) therapeutic efficacy trial protocol: patients with P. falciparum monoinfection, febrile or with a history of fever, asexual parasite density >500/μl and <100 000/μl and willingness to consent to follow-up (World Health Organization 2009). We excluded pregnant patients and those with signs of severe malaria. The study population represents a cross-section of the patients who presented to the local clinic or were recruited through active case detection in nearby communities.

image

Figure 1. Sentinel sites of the National Antimalarial Drug Resistance Monitoring System by year, parasite species, and state (district), India, 2009–2010.

Download figure to PowerPoint

Data collection

The data collection methods have been previously described (Mishra et al. 2012). Briefly, clinical and demographic information was recorded from each patient at enrolment. Patients were followed during the first 3 days of treatment (artesunate + sulfadoxine–pyrimethamine given orally, 4 mg/kg for 3 days + 25/1.25 mg/kg as per national guidelines) and then at weekly intervals from enrolment until day 28. At each follow-up visit, a physical exam was conducted, and thick and thin blood smears were prepared. Using routine oil immersion reading at 100 × on Giemsa-stained thick smears, parasites were counted until 200 white blood cells (WBC) if gametocytes and asexual stages were present or until 1 000 WBC to declare a slide negative for both. A count of 8 000 WBC/μl of blood was assumed to obtain a final density for asexual and sexual parasites. The data were double entered into the WHO therapeutic efficacy database, and blood slides were cross-checked by expert microscopists.

Case and predictor definitions

We defined gametocytemia as the presence of gametocytes in the peripheral blood smear, at any visit between day 0 and day 2 of follow-up, in a patient eligible for a therapeutic efficacy study in 2009 or 2010. The half-life of mature gametocytes is estimated at 4 to 6 days once in peripheral blood (Bousema et al. 2010). Thus, any peripheral blood gametocytemia up to day 2 is accurate for measuring risk factors. Gametocytemia detected later in follow-up, however, may have different origins. For example, gametocytogenesis from persisting asexual stages, reinfection or the release of sequestered developing gametocytes could explain any later onset of gametocytemia.

We selected predictors associated with gametocytemia in prior literature that could be feasibly identified during routine curative care: age, sex, region, previous antimalarial drug intake, current fever, history of fever, season and asexual parasite density (Bousema & Drakeley 2011). The recall period for history of fever was the past 48 h and for taking an antimalarial drug was the past week. Fever was defined as axillary temperature ≥37.5 °C at the time of enrolment. We designated region using geographic clusters associated with different malaria ecotypes: western India as Gujarat, Mumbai and Rajasthan, central India as Andhra Pradesh, Chhattisgarh, Gadchiroli, Jharkhand, Madhya Pradesh and Orissa, and northeast India as the Assam, Meghalaya and West Bengal sites (Sharma 1999; Sonal et al. 2010). We classified season by month of enrolment: monsoon–June-August, post-monsoon–September-November and winter–December and January.

Data analysis

The analysis included all patients who completed 2 follow-up visits after enrolment. Missing gametocytemia data, due to withdrawal and loss to follow-up during the 3-day treatment phase, constituted less than 3% of the overall sample. None of the predictors, age, sex, region, previous antimalarial drug intake, current fever, history of fever, season and asexual parasite density, had missing data. We calculated the prevalence of gametocytemia among levels of predictors and in the overall population. We also calculated the proportion of parasites that were gametocytes by dividing the enrolment gametocytemia by the sum of the enrolment gametocyte and asexual parasite densities. To calculate the proportion of the reservoir for potential transmission in each age category, we first multiplied the geometric mean of the maximum gametocyte density, as well as the geometric mean of the average gametocyte density for sensitivity analysis, during day 0 through day 2 in the age category by the number of gametocytemic individuals in the age category. Second, we summed the gametocyte load of each age category to obtain the population gametocyte load. Third, we divided the gametocyte load of each age category by the total gametocyte load leading to the proportion of the reservoir for potential transmission represented by the age category. Finally, we also calculated the unweighted proportions by age category without including gametocyte density assuming transmission is not density dependent when gametocytemia is microscopically detectable. The use of transmission is qualified with the term ‘potential’ as our approach ignores differences in gametocyte infectivity (due to immunity) and vector biting rates (due to uncovered body surface area, etc.) between the age groups.

Unconditional multivariate logistic regression was employed to build a reference predictive model between the demographic and clinical factors and gametocytemia. To account for the clustering of data in each trial, we estimated cluster robust standard errors with district as the unit (Eldridge & Kerry 2012). We estimated the crude prevalence odds ratio and its 95% confidence interval between each predictor and gametocytemia. We included all predictors associated with gametocytemia (P value < 0.25 to avoid the exclusion of important variables) in those bivariate analyses. We assessed collinearity between each pair of predictive factors by calculating the odds ratio and selected among collinear variables (odds ratio >3) based on their substantive value. We evaluated two-way interactions between all pairs of predictors and retained all product terms with P value < 0.25.

To simplify field use of the algorithm, we examined reduced models that had similar predictive power and adequate fit compared with the reference model. We used backwards elimination using the Wald test to remove predictors with P value < 0.10 starting with interaction terms and proceeding to the variable with the highest P value. We assessed the predictive power of models reduced by removing variables or collapsing across categories, through comparison of the area under the receiver operating characteristic (ROC) curve. We evaluated model fit using the Hosmer–Lemeshow test (P value > 0.1) (Hosmer & Lemeshow 2000). Then a scoring system was created from the logistic model output using the regression coefficient, to preserve the multiplicative nature of the score, for each predictor in the reduced model. We multiplied the regression coefficient by 10 and rounded to the nearest integer to simplify score use (Harrell 1996). A final score for each patient was obtained by summing the individual scores from their predictor values. To determine the utility of the scoring system, we evaluated the sensitivity, specificity, false negatives, false positives and the area under the ROC curve of different score cut-points. We calculated false negatives using the formula.

(1–sensitivity) * gametocytemia prevalence * N and false positives as (1 – specificity) * (1–gametocytemia prevalence) * N.

We also calculated the per cent of the population correctly classified, and the per cent of the population that would be treated if scores were used to target gametocytocidal therapy. We imported the final data set into STATA (v10) and used it for all analyses.

Study power

Assuming a gametocytemia prevalence of at least 10% and α = 0.25, we estimated more than 95% power in the study to detect risk factors prevalent among at least 7.5% of controls with a prevalence odds ratio of 2 or more.

Ethical clearance

The Scientific Advisory Committee of the National Institute of Malaria Research approved the original trials, and the Institutional Review Board of the University of North Carolina approved the secondary analysis study.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

In 2009 and 2010, 1 372 patients with P. falciparum malaria were recruited into therapeutic efficacy trials of antimalarial drugs. Among these patients, 19 voluntarily withdrew, 3 received outside treatment, 2 contracted other illnesses and 9 were lost to follow-up. After removing 4 patients who were missing gametocytemia data, our complete case population was 1 335. The majority of the study population was, independently, adult, male, from central India, and enrolled in the post-monsoon (Table 1). The proportion of patients with gametocytemia on day 0, day 1 and day 2 was 13% (n = 179), 15% (n = 201) and 15% (n = 203), respectively. Overall, the prevalence of gametocytemia, that is, gametocytes detected in blood films on any day from day 0 through day 2 was 19% (n = 248); this varied in relation to demographic and clinical classifications (Table 1). In the unadjusted bivariate associations, gametocytemia decreased with both increasing age and parasite density categories, while it was lower among those without fever at enrolment or a history of fever prior to enrolment. Men and patients who reported fever or unknown previous antimalarial intake also had a higher prevalence of gametocytemia. The proportion of malaria patients with gametocytemia varied by region and fell along a western to eastern India axis.

Table 1. Prevalence of gametocytemia in relation to demographic and clinical factors of patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010
CharacteristicGametocytemiaTotal
n Row% n Col%
  1. PAI, previous antimalarial intake; PD, asexual parasite density; POR, prevalence odds ratio; CI, confidence interval; Col, column.

Age (years)
<528271058
5–141012050338
≥151191672754
Sex
Male1572077358
Female911656242
PAI
No241181 31098
Yes/unknown728252
Region
Central841173155
Western1534137128
Northeast11523317
Fever day 0
Yes1441692970
No1042640630
Season
Monsoon34840630
Post-monsoon1742862046
Winter401330923
PD (#/μl)
<5 0001182548036
5 000–49 9991171674656
≥50 00013121098
History of fever
Yes237181 31799
No1161181

The unadjusted prevalence of gametocytemia ranged from 26% (n = 103) among ages 1–4 years to 14% (n = 96) in those aged 50 years or older (Figure 2). Inversely, the proportion of the total parasite population consisting of gametocytes increased with age from 3% in 1–4-year olds to 8% in people aged 50 or older. The average densities, represented by the geometric mean of the maximum gametocytemia and mean gametocytemia from day 0 through day 2, were 117 and 66 gametocytes/μl, respectively. The density of gametocytes was higher in children than adults (Table 2), which was similar to the trend observed with asexual parasite density at enrolment (data not shown). In unadjusted analysis, gametocyte densities were similar in western and central India but higher in northeast India in all age categories (Table 2). Adults (age 15 years or more), who were 54% of the study population and among whom 16% carried gametocytes, constituted approximately 44% of the reservoir for potential transmission. School-age children (age 5–15 years), who were 38% of the study population and among whom 20% carried gametocytes, constituted approximately 44% of the reservoir for potential transmission. Young children (age less than 5 years), who were 8% of the study population and among who 27% carried gametocytes, constituted approximately 12% of the reservoir for potential transmission. These estimates did not differ by region except for northeast India where the relative contributions of school-age children and younger children were reversed compared with other regions. These estimates also did not differ whether the maximum or mean gametocyte density was used. Assuming transmission is not gametocyte density dependent, the unweighted contribution for adults towards potential transmission increased in the total population (Table 2).

Table 2. The contribution of age groups to the reservoir for potential transmission using the unweighted, maximum, or mean, day 0 to 2 gametocyte density in patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010
AreaAgeNPGMaximumMean
GDIRGDIRUW
  1. PG, prevalence of gametocytemia; GD, geometric mean gametocyte density per microlitre; IR, proportion of the reservoir for potential transmission; UW, unweighted IR without accounting for gametocyte density.

Central<56420870.11460.110.15
5–14301111410.43770.450.39
≥15366101300.46650.440.45
Western<513771520.09240.040.07
5–14136491220.49470.500.43
≥1522235880.42370.460.50
Northeast<528182080.361530.460.45
5–146631200.08840.100.18
≥1513934020.561820.440.36
Total<5105271240.12690.120.11
5–14503201280.45710.450.41
≥15727161020.43580.430.48
Table 3. Adjusted prevalence odds ratios in the reference and final models, regression coefficients and risk scores for predicting gametocytemia in patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010
VariableReference model OR (95%CI) AUC = 0.766Final model OR (95%CI) AUC = 0.762Logistic regression coefficientRisk score
  1. a

    For the entire age category.

  2. PAI, previous antimalarial intake; PD, asexual parasite density; OR, odds ratio; CI, confidence interval.

Age (years), Sex
<5, male2.341.03, 5.283.88a2.13, 7.061.3614
<5, female6.883.05, 15.6    
5–14, male1.300.79, 2.141.51a0.96, 2.360.414
5–14, female2.071.18, 3.64    
≥15, male2.071.26, 3.401.00a   
≥15, female1.00     
Sex
Male 1.491.06, 2.100.404
Female  1.00   
PAI
No1.00     
Yes/unknown1.691.00, 2.871.670.99, 2.810.515
Region
Central2.981.69, 5.282.771.63, 4.691.0210
Western16.39.44, 28.117.19.98, 29.32.8428
Northeast1.00     
Fever day 0
Yes1.300.93, 1.81   
No1.00     
PD (#/μl)
<5 0001.540.74, 3.24   
5 000–49 9991.440.72, 2.88    
≥50 0001.00     
image

Figure 2. Prevalence of gametocytemia and the per cent of total parasites that were gametocytes by age category of patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010.

Download figure to PowerPoint

Age, sex, the age–sex product interaction, region, previous antimalarial intake, fever at enrolment and parasite density category remained in the reference model (Table 2). In the simplified model, age, sex, region, and previous antimalarial intake alone provided similar predictive ability and model fit (P value = 0.32) (Table S1). Possible risk scores ranged from 0 to 65 although the minimum and maximum scores were 0 and 45. The median risk score was 14 (interquartile range: 10, 28). Residing in the western region was the highest scoring predictor with 28 points; age 5–14 years and male sex were the lowest scoring predictors with 4 points each (Table 3). No cut-off point yielded a sensitivity above 75% and a specificity below 75%. For example, if the goal of a control programme was to treat at least 90% of gametocyte carriers, a risk score of 14 or more provided 91% (95%CI: 88, 95) sensitivity and 33% (95%CI: 31, 36) specificity (Table 4). Applied in our study population of 1 335 patients of whom 248 were gametocytemic, 71% of the population would receive treatment with 22 false negatives and 723 false positives. The area under the ROC curve for predicting gametocytemia was 0.76 (95%CI: 0.73, 0.80) (Figure 3). For comparison, the AUC of the model using all predictors was 0.79 with two-way interactions and 0.82 with all possible interactions.

Table 4. Performance of different risk score cut-offs for detecting gametocytemia in patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010
Score ≥SensitivitySpecificityNumber of FNNumber of FPPer cent treated
  1. FN, false negative; FP, false positive.

01.000.0001087100
41.000.031105998
80.980.15592487
100.980.18688885
130.920.322073872
140.910.332272371
150.760.595944147
170.760.606043547
180.750.616342746
190.670.758127733
230.670.758227033
270.630.789124230
280.620.809521928
320.570.8510716523
330.190.97200326
360.190.97200316
370.061.0023351
410.051.0023531
450.021.0024420
460.001.0024800
image

Figure 3. Receiver operator characteristics curve with risk score cut-points for predicting gametocytemia in patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We observed a high prevalence of gametocytemia in India, and adults constituted a substantial proportion of the reservoir for potential transmission in our sampled population. While a predictive model for gametocytemia identified several easily screened risk factors, the ability of the clinical algorithm to sensitively and specifically detect gametocytemia was low.

We observed a higher prevalence of gametocytemia than previously reported although there were large, albeit unadjusted, regional differences. Gametocytemia was most common in western India, which is composed of two distinct ecotypes: rural, low-transmission malaria, and urban slum and migrant-labour associated malaria. Both ecotypes could be associated with a high risk of gametocytemia through low immunity and/or poor access to quality care, especially for migrants. Central India and northeast India include higher transmission areas with forest-associated malaria albeit through different vectors. In the northeast, access to care is better and the use of artemisinin combination therapies began earlier, which may explain the lower prevalence of gametocytemia relative to central India (Sonal et al. 2010).

The reservoir for potential transmission in our study population was distributed throughout the age spectrum. Traditionally, children were thought to be the primary reservoir for transmission (Drakeley et al. 2006). Young and school-age children did contribute to the reservoir for potential transmission disproportionate to their population because of their higher prevalence of gametocytemia and higher gametocyte densities. Still, adults constituted nearly half of the potential reservoir of infection simply due to their larger population. In adults, a higher proportion of the parasite population was gametocytes although this was largely due to a smaller denominator as the asexual parasite density decreased with age (data not shown). The contribution of adults in malaria transmission may be higher than the potential reservoir estimated by us when accounting for other factors such as their larger surface area for biting (Port et al. 1980). These results underscore the need to examine absolute measures of frequency rather than relative measures to inform public health conclusions. While children may be at higher risk of infection and may individually contribute more towards transmission, most malaria patients in India and other low to moderate transmission settings are adults. Many malaria intervention strategies, such as school-based screening or preferential net distribution in mass campaigns, target children. Given the vulnerability of children, this focus is not inappropriate, but without addressing the burden in adults, we cannot maximise malaria control.

Four variables (region, age, sex and previous antimalarial drug intake) predicted gametocytemia in our model. Previous intake of non-gametocytocidal antimalarial drugs is thought to induce ‘stress’ on the parasite, which activates gametocytogenesis (Butcher 1997). Age, sex and region would, presumably, be associated with gametocytemia through one of two mechanisms: (i) immunity, primarily determined by transmission intensity and the exposure of specific risk groups, and (ii) treatment access or treatment seeking behaviour as gametocytemia increases with longer durations of infection (Doolan et al. 2009). Parasite density and fever at the time of enrolment, which were removed from the final model, are also distal effects of, rather than proximal markers of, the aforementioned mechanisms, which may explain their inability to predict gametocytemia in our model.

The use of a predictive model to detect gametocytemia generated an algorithm that ranked a positive case selected at random from our study population higher than a negative case selected at random 76% of the time (AUC). No cut-off point yielded an acceptable sensitivity (>95%) and specificity (>90%) according to criteria developed for malaria rapid diagnostic tests (Bell & Peeling 2006; WHO, Publications on rapid diagnostic tests). As an illustration, if we selected 90% sensitivity or more as a desirable criterion, we could only achieve 33% specificity. We also did not validate our algorithm on independent data, and hence, its performance in our study population could be considered a best-case scenario. Other strategies for selecting a predictive model are unlikely to produce better clinical algorithms with the available data as the AUC of the final model was close to the AUC of the saturated model. Alternative data, however, could produce better clinical algorithms if other easily measured predictors existed. While our performance would not suffice for a disease diagnostic, one could argue the direct costs of using an algorithmic approach are non-existent; so any reduction in false positives is a benefit compared with universal treatment. However, substantial indirect costs may exist. Considerations of implementing any clinical algorithm must account for the operational challenges in individual level targeting including the costs of training, the time required for patient assessment and increased programme complexity. Poor prospects for future improvement in model performance coupled with the likelihood of considerable indirect costs of implementation suggest that a clinical predictive approach for targeting gametocytemia is not viable.

Our study had several limitations. We used microscopy for the measurement of gametocytemia, which is less sensitive than molecular techniques. However, in studies comparing the two methods, the latter increased the magnitude of gametocytemia but did not alter its age structure, circulation time and other trends (Bousema et al. 2010; Ouédraogo et al. 2010). Interpreting the functional relevance of submicroscopic gametocytemia is also difficult. While submicroscopic density infections can infect mosquitoes, the probability of infection, the proportion of mosquitoes infected and the density of infection in mosquitoes are positively correlated with gametocyte density (Schneider et al. 2007; Ouédraogo et al. 2009). Next, we completed enrolment at each site over 1–2 months, which restricted the analysis of seasonal trends of gametocytemia. Our population cross-section was also not representative of the population at risk. It was representative of the population encountered by the control programme through active and passive case detection. Thus, we could not assess the contribution of asymptomatic carriers to transmission in thus study but, at present, there may also be no valid means to do so (Laishram et al. 2012). We also were not able to assess gametocytemia in patients with mixed infections. Interspecies dynamics may alter gametocytogenesis; for example, epidemiological evidence for P. falciparum gametocytemia as an indicator for occult mixed infection exists (Lin et al. 2011). Finally, we used the presence of gametocytes in peripheral blood as a proxy for infectiousness. In reality, infectivity is modified by a number of factors; it can be assessed most directly through membrane-feeding experiments, but these are labour intensive and would not be possible in a large survey needed for generalisable results (Awono-Ambene et al. 2001).

In a population of P. falciparum patients from a national network of sentinel sites, we conclude gametocytemia was common, adults were an important component of the reservoir for potential transmission, and clinical algorithms based on predictive modelling were not effective for the detection of gametocytemia. Due to the wide age distribution of gametocytemia, and the difficulty of targeting using clinical prediction, we recommend universal application, if any, of gametocytocidal interventions among confirmed malaria patients. Future research on gametocytemia should prioritise the measurement of the asymptomatic reservoir, conduct longitudinal assessments and validate gametocytemia as an indicator for treatment access.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

NKS was supported by a fellowship from the Paul and Daisy Soros Foundation and National Institutes of Health Medical Scientist Training Program grant. This paper was approved by the NIMR publications screening committee (approval no. 026/2012).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Awono-Ambene HP, Diawara L & Robert V (2001) Comparison of direct and membrane feeding methods to infect Anopheles arabiensis with Plasmodium falciparum. American Journal of Tropical Medicine and Hygiene 64, 3234.
  • Bell D & Peeling RW (2006) Evaluation of rapid diagnostic tests: malaria. Nature Reviews Microbiology 4 (Suppl. 9), S34S38, doi:10.1038/nrmicro1524.
  • Bousema T & Drakeley C (2011) Epidemiology and infectivity of Plasmodium falciparum and Plasmodium vivax gametocytes in relation to malaria control and elimination. Clinical Microbiology Reviews 24, 377410. doi:10.1128/CMR.00051-10.
  • Bousema T, Okell L, Shekalaghe S et al. (2010) Revisiting the circulation time of Plasmodium falciparum gametocytes: molecular detection methods to estimate the duration of gametocyte carriage and the effect of gametocytocidal drugs. Malaria Journal 9, 136. doi:10.1186/1475-2875-9-136.
  • Butcher GA (1997) Antimalarial drugs and the mosquito transmission of Plasmodium. International Journal for Parasitology 27, 975987.
  • Doolan DL, Dobaño C & Baird JK (2009) Acquired Immunity to Malaria. Clinical Microbiology Reviews 22, 1336. doi:10.1128/CMR.00025-08.
  • Drakeley C, Sutherland C, Bousema JT, Sauerwein RW & Targett GAT (2006) The epidemiology of Plasmodium falciparum gametocytes: weapons of mass dispersion. Trends Parasitol 22, 424430. doi:10.1016/j.pt.2006.07.001.
  • Eldridge S & Kerry S (2012) A Practical Guide to Cluster Randomised Trials in Health Services Research. John Wiley & Sons, West Sussex, UK, pp. 299.
  • Harrell F (1996) Regression coefficients and scoring rules. Journal of Clinical Epidemiology 49, 819.
  • Hosmer DW & Lemeshow S (2000) Applied Logistic Regression, John Wiley & Sons, Hoboken, NJ, pp.397
  • Kar PK, Dua VK, Gupta NC, Gupta A & Dash AP (2009) Plasmodium falciparum gametocytaemia with chloroquine chemotherapy in persistent malaria in an endemic area of India. Indian Journal of Medical Research 129, 299304.
  • Karl S, Davis TME & St-Pierre TG (2009) A comparison of the sensitivities of detection of Plasmodium falciparum gametocytes by magnetic fractionation, thick blood film microscopy, and RT-PCR. Malaria Journal 8, 98. doi:10.1186/1475-2875-8-98.
  • Kumar A, Valecha N, Jain T & Dash AP (2007) Burden of malaria in India: retrospective and prospective view. American Journal of Tropical Medicine and Hygiene 77, 6978.
  • Laishram DD, Sutton PL, Nanda N et al. (2012) The complexities of malaria disease manifestations with a focus on asymptomatic malaria. Malaria Journal 11, 29. doi:10.1186/1475-2875-11-29.
  • Lin JT, Bethell D, Tyner SD et al. (2011) Plasmodium falciparum gametocyte carriage is associated with subsequent Plasmodium vivax relapse after treatment. PLoS ONE 6, e18716. doi:10.1371/journal.pone.0018716.
  • Macdonald G (1956) Theory of the eradication of malaria. Bulletin of the World Health Organization 15, 369387.
  • Mishra N, Singh J, Srivastava B et al. (2012) Monitoring antimalarial resistance in India via sentinel sites: outcomes and risk factors for treatment failure, 2009–2010. Bulletin of the World Health Organization 91, 119.
  • Mlambo G, Vasquez Y, LeBlanc R, Sullivan D & Kumar N (2008) A filter paper method for the detection of Plasmodium falciparum gametocytes by reverse transcription polymerase chain reaction. American Journal of Tropical Medicine and Hygiene 78, 114116.
  • Mohapatra SS, Govardhini P, Jambulingam P & Pani SP (1992) Some observations on Plasmodium falciparum gametocytaemia in natural infections in an endemic area of Koraput district, Orissa. Indian Journal of Malariology 29, 6972.
  • Mohapatra PK, Prakash A, Bhattacharyya DR & Mahanta J (1998) Epidemiological importance of younger age group during malaria epidemic in PHC Tamulpur, Assam. Journal of Communicable Diseases 30, 229232.
  • NVBDCP, National Vector Borne Disease Control Programme (n.d.). Available: http://nvbdcp.gov.in/malaria2.html. Accessed 6 February 2012.
  • Ouédraogo AL, Bousema T, Schneider P et al. (2009) Substantial contribution of submicroscopical Plasmodium falciparum gametocyte carriage to the infectious reservoir in an area of seasonal transmission. PLoS ONE 4, e8410. doi:10.1371/journal.pone.0008410.
  • Ouédraogo AL, Bousema T, De Vlas SJ et al. (2010) The plasticity of Plasmodium falciparum gametocytaemia in relation to age in Burkina Faso. Malaria Journal 9, 281. doi:10.1186/1475-2875-9-281.
  • Port GR, Boreham PFL & Bryan JH (1980) The relationship of host size to feeding by mosquitoes of the anopheles gambiae giles complex (Diptera: Culicidae). Bulletin of Entomological Research 70, 133144. doi:10.1017/S0007485300009834.
  • Rajagopalan PK, Das PK, Pani SP et al. (1990) Parasitological aspects of malaria persistence in Koraput district Orissa, India. Indian Journal of Medical Research 91, 4451.
  • Schneider P, Schoone G, Schallig H et al. (2004) Quantification of Plasmodium falciparum gametocytes in differential stages of development by quantitative nucleic acid sequence-based amplification. Molecular and Biochemical Parasitology 137, 3541. doi:10.1016/j.molbiopara.2004.03.018.
  • Schneider P, Bousema JT, Gouagna LC et al. (2007) Submicroscopic Plasmodium falciparum Gametocyte Densities Frequently Result in Mosquito Infection. The American Journal of Tropical Medicine and Hygiene 76, 470474.
  • Sharma VP (1999) Current scenario of malaria in India. Parassitologia 41, 349353.
  • Singh V, Mishra N, Awasthi G, Dash AP & Das A (2009) Why is it important to study malaria epidemiology in India? Trends in Parasitol 25, 452457. doi:10.1016/j.pt.2009.06.004.
  • Sinton J (1926) Studies in malaria, with special reference to treatment. Part II. The effects of treatment on the prevention of relapse in infections with plasmodium falciparum. Indian Journal of Medical Research 13, 895916.
  • Sonal GS, Thakor HG, Joshi C, Arora P, Gupta RK, Dhariwal AC (2010) Epidemiological status of malaria and scaling up of interventions in India. Journal of the Indian Medical Association 108, 840843.
  • Taylor LH & Read AF (1997) Why so few transmission stages? Reproductive restraint by malaria parasites. Parasitol Today (Regul Ed) 13, 135140.
  • WHO, Publications on rapid diagnostic tests (n.d.). WHO. Available: http://www.who.int/malaria/publications/rapid-diagnostic-tests/en/index.html. Accessed 4 May 2012.
  • World Health Organization (2009) Methods for surveillance of antimalarial drug efficacy.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
tmi12119-sup-0001-TableS1.docxWord document17KTable S1. Cross-tabulation according to predictors identified in the reduced model of patients from the National Antimalarial Drug Resistance Monitoring System, India, 2009–2010

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.