Anopheles; Culicidae; malaria; Plasmodium falciparum

Transcription

1 American Journal of Epidemiology Copyright 1997 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 145, No. 10 Printed in U.S.A Dose- and Time-dependent Relations between Infective Anopheles Inoculation and Outcomes of Plasmodium falciparum Parasitemia among Children in Western Kenya Peter D. McElroy, 1 John C. Beier, 34 Charles N. Oster, 3-5 Fred K. Onyango, 3 Aggrey J. Oloo, 6 Xihong Lin, 7 Christine Beadle, 1 and Stephen L. Hoffman 1 Blood-stage level Plasmodium falciparum infection (parasitemia density) is generally elevated prior to, or at the time of, clinical presentation of severe pediatric malaria episodes. Intensity of exposure to infective Anopheles mosquito bites is a suspected determinant of higher density parasitemia. Analyses of entomologic and parasitologic data collected in were conducted to investigate whether the dose of infective bites predicted the incidence or degree of P. falciparum parasitemia in Kenyan children <6 years old. At 1 consecutive 30-day intervals, a new cohort (n «each) was enrolled, cured of malaria parasites, and monitored over 84 days for recurrent parasitemia. Outcomes included time to parasitemia, time to parasitemia s5,000/pj, and parasitemia density. Ecologic and individual-level analyses were conducted. The mean infective bite exposure experienced by each cohort was significantly associated with the incidence of parasitemia (age-adjusted r = 0.38, p = 0.0) and more strongly associated with the incidence of parasitemia z5,000/fi\ (age-adjusted r = 0.7, p < 0.001). The infective bite dose, analyzed as a time-dependent covariate, was associated with a.8 times higher rate of parasitemia >5,000/*il among children exposed to >1 infective bite per day as compared with the referent (rate ratio (RR) =.8, 95% confidence interval (Cl) ). Cumulative infective bite exposure, exposure duration, and age were significant predictors of recurrent parasitemia density in multiple linear regression analyses. The results support the contention that reductions in P. falciparum transmission intensity, in the absence of complete elimination, will reduce higher level parasitemia among African children. Am J Epidemiol 1997;145: Anopheles; Culicidae; malaria; Plasmodium falciparum The global malaria situation is intensifying. Over two billion people now reside in areas where Plasmodium falciparum is transmitted, and each year million new infections with this parasite result in million deaths (1). An important goal of ma- Received for publication May 8,1996, and accepted for publication January 8,1997. Abbreviations: Cl, confidence interval; RR, rate ratio. 1 Malaria Program, Naval Medical Research Institute, Bethesda, MD. Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Ml. 3 US Army Medical Research Unit, Nairobi and Kisumu, Kenya. 4 Department of Tropical Medicine, School of Public Health, Tulane University, New Orleans, LA. 5 Department of Infectious Diseases, Walter Reed Army Institute of Research, Washington, DC. 6 Kenya Medical Research Institute, Vector Biology and Control Research Centre, Kisumu, Kenya. 7 Department of Biostatistics, School of Public Health, University of Michigan, Ann Arbor, Ml. Reprint requests to Dr. Stephen L Hoffman, Malaria Program, Naval Medical Research Institute, 1300 Washington Ave., Rockville, MD 085. This paper was prepared under the auspices of the US Government and is therefore not subject to copyright. laria control is to reduce the incidence of severe malarial anemia and cerebral malaria, two disease outcomes accompanied by substantial mortality among young African children. However, the underlying biologic mechanisms of these serious manifestations are quite contentious, and a clear consensus regarding predictors of disease has yet to emerge (-4). One documented risk factor of clinical severity is elevated numbers of asexual erythrocytic-stage P. falciparum parasites in either the peripheral blood circulation (parasitemia) or sequestered in the deep vasculature (5-9). In holoendemic malarious areas, innate and acquired immunity influences P. falciparum parasitemia density (number of asexual parasites per /il of blood). Other host factors associated with malarial illness include age, hemoglobinopathies, history of chronic infection, nutrition, and use of antimalarials (10, 11). How these variables are influenced by transmission intensity of P. falciparum, specifically the dose of parasite inoculum, has not been sufficiently investigated under natural transmission conditions. Insecticide-impregnated 945

2 946 McElroy et al. bednet trials provide indirect evidence that intensity of exposure to sporozoite inoculation (the parasite stage directly inoculated into humans by the bite of an infective Anopheles mosquito) may be a determinant of disease severity (1, 13). The incidence of clinical malaria declined in villages randomized to bednets, but the prevalence and incidence of P. falciparum infection often remained comparable to those of control villages (14-16). Other studies have examined whether the inoculum dose is predictive of the incubation period or density of parasitemia, but those data were obtained from experimentally induced infection in nonimmune adults exposed to homologous parasite phenotypes (17-0). The effectiveness of protective immune responses that limit disease severity may be partially dependent upon the number of pre-erythrocytic-stage parasites arising per inoculation and the duration of time between inoculations (). Sporozoites and hepatic-stage parasites are targets of humoral and cellular immune mechanisms that destroy parasites prior to the erythrocytic stage of infection where clinical illness develops (1). In some African regions, exposure to sporozoite inoculation may be as frequent as 0-1,000 times per annum (, 3). Inhabitants of less endemic areas experience only a fraction of such exposure (4, 5). In a -year prospective study ( ), detailed entomologic and parasitologic data were collected from over 1,000 young children in western Kenya. We have previously reported a significant association between the level of exposure to infective anopheline bites over the 8-day period prior to enrollment and the density of P. falciparum parasitemias among these children (6). In addition, the incidence of recurrent parasitemia and the density of recurrent parasitemia following drug treatment were both associated with seasonal transmission intensity (7, 8). However, none of these earlier reports attempted to address the dynamic relations among a continuum of sporozoite dose, duration of exposure, age, level of parasitemia at enrollment, village of residence, sex, and recurrent P. falciparum parasitemia over the entire -year observation period. We have explored this complex interdependence and report evidence of a causal relation between the level of infective bite dose and parasitemia density among children in western Kenya. MATERIALS AND METHODS Study site and participants The study was conducted in six contiguous villages (Got Onyango, Miyare, Saradidi, Abwao, Rariw, and Luoro) on the northeastern shore of Lake Victoria in western Kenya. The 1985 population was 8,4 persons, living in 1,056 houses dispersed over 10.5 km. Malaria in this region is holoendemic, with perennial transmission of P. falciparum. The intensity of exposure to sporozoite inoculation fluctuates, with the highest transmission occurring between April and July. During peak transmission, exposures of five infective anopheline bites per person per night have been detected (7). Previous enzyme-linked immunosorbent assay data have indicated P. falciparum sporozoites in 95 percent of the infective anopheline mosquitoes collected in western Kenya (9). Plasmodium ovale and Plasmodium malariae are also transmitted at this site, but neither significantly contributes to the overall burden of disease or death. All 6-month- to 6-year-old children free of serious underlying diseases and whose parents provided informed, written consent were eligible to participate. As previously reported, 94.4 percent of the original 1,007 study participants were infected with P. falciparum upon enrollment (6). Study design and procedures Households with children under 6 years of age were selected and randomized to 1 groups. After 1 months, further selection and randomization of households provided nine additional groups, for a total of 1 study cohorts. Enrollment of the first cohort of children occurred in January Subsequently, every fourth week through July 1987 a new cohort of approximately children was enrolled. Upon enrollment (day 0), a physician evaluated each child, and a technician obtained a blood film for parasite identification. Each child, regardless of parasitemia status, received 5 mg/kg of sulfadoxine and 1.5 mg/kg of pyrimethamine to eliminate blood-stage infection. Children were visited daily by a village health worker and referred to a clinic for examination and blood film if illness was present. Blood films were obtained from all children at 7, 14, 8, 4, 56, 70, and 84 days post-sulfadoxine/pyrimethamine therapy. Members of each cohort were followed for a maximum of 84 days. Thick blood films were Giemsa stained and independently examined for P. falciparum by two expert microscopists. Each microscopist examined 00 fields (X 1,000) before characterizing a specimen as "negative." P. falciparum parasitemia density was derived using the number of parasites/00 leukocytes and an assumed leukocyte density of 8,000 cells//xl of blood (30). The lower limit of parasite detection was four asexual forms//xl, assuming that 00 thick film fields equal 0.5 /xl of blood (31). The time to and density of the first recurrent parasitemia following sulfadoxme/pyrimethamine treatment were obtained from the blood film data. Because of insufficient

3 Plasmodium falciparvm Transmission in Kenya 947 day 0 sulfadoxine/pyrimethamine dosage for the first cohort, it was excluded from all analyses. During the -month study period, the intensity of P. falciparum transmission was estimated using a well-established malariometric parameter: the entomologic inoculation rate. This parameter expresses, at the community level, the intensity of human exposure to blood meal-seeking, sporozoite-infected, anopheline mosquitoes. The methods used for the weekly mosquito collections have been described (). Briefly, two men trained in mosquito collection were assigned one night per week to a representative household in each village. The same six houses were used each week, but mosquito collectors were rotated. The two men collected blood meal-seeking mosquitoes from one another's exposed legs for 30-minute intervals, followed by 30-minute rest periods, from sunset to sunrise (approximately 6:30 p.m. to 6:00 a.m.). All collected mosquitoes were delivered to a laboratory in Kisumu the following morning, separated according to species, and dissected to detect sporozoites in the salivary glands. The daily entomologic inoculation rate (EIR) was estimated using the formula: EIR = (human-biting rate) X no. of anopheline mosquitoes with salivary gland sporozoites total no. of dissected anopheline mosquitoes' The mean daily entomologic inoculation rate was estimated for 44 consecutive -week intervals during the study (figure 1). The months of highest transmission corresponded with the peak seasonal rainfall (7). Outcome variables Incidence of first recurrent parasitemia. The P. falciparum incidence rates described in an earlier report were based on a denominator containing individuals with negative blood films on day 8 (7). Individuals with positive films on day 8 were excluded from the denominator to minimize the chance that a recurrent parasitemia was due to sulfadoxine/pyrimethamine drug resistance. However, on the basis of further analyses comparing cohorts exposed to the highest and lowest transmission intensities, we concluded that most positive films on day 8 (n = 93) were new infections, not recrudescent infections (8). Thus, the incident cases of first recurrent parasitemia in these current analyses included only children with negative day 14 blood films. The person-time at risk was calculated from day 14 through the day of first detectable recurrent P. falciparum parasitemia (figure ). Children lost to follow-up or who never developed a recurrent parasitemia contributed person-time from day 14 through the date of their last available negative blood film. Incidence of first recurrent high-density parasitemia. The first recurrent P. falciparum parasitemias ^5,000/jxl were characterized as high-density parasitemias. Selection of ^5,000//xl as an endpoint was based on recent studies where the likelihood of a clinical malaria episode was significantly increased when the density exceeded 5,000//A1 (8, 3, 33). Density of first recurrent parasitemia. The density of first recurrent P. falciparum parasitemia was also Two-week interval (Feb Oct. 1987) FIGURE 1. Mean daily entomologic inoculation rates (EIRs) for 44 consecutive -week intervals at the western Kenya study site. Exact dates on the abscissa correspond to dates of cohort enrollment (day 0). Each point represents the mean entomologic inoculation rate for the previous -week period.

4 948 McElroy et al. Day 0 14 I 8 - duration of exposureperson-time I 4 I FIGURE. Timeframe for all study participants, western Kenya, Included are the fortnightly intervals of blood film acquisition between enrollment (day 0) and cessation of cohort follow-up (day 84). Also included is a representation of how the duration of exposure to sporozoite inoculation and the accumulated person-time at risk were derived for a child (x) with first recurrent parasitemia detected on day 70. Sex. In our earlier analyses, the child's sex was not associated with prevalence or density of blood-stage infection on the day of enrollment (6). However, it was of interest to determine whether sex was associated with the incidence or density of recurrent parasitemia. Village of residence. Each child's village of residence was included in all analyses to account for focal differences in transmission intensity. Fewer mosquitoes were generally captured in Got Onyango village; hence, this was considered the referent group. investigated as a continuous variable. All parasite density values were log transformed prior to analysis. Exposure variables Duration of infective bite exposure. In a human, 8-14 days generally transpire between P. falciparum sporozoite inoculation by a mosquito and the time of a first detectable parasitemia. The duration of exposure to sporozoite inoculation was thus derived as die period starting on day 0 and ending 14 days immediately prior to a child's first recurrent positive blood film. For example, in a child with a first recurrent parasitemia on day 70, duration of exposure to inoculation was 56 days (figure ). During the 84-day follow-up period, the maximum duration of exposure for any child was 70 days. Dose of infective bites. Two measures of infective bite dose were derived for each child. Each -week estimate of the daily mean dose of infective bite exposure experienced by a child (figure 1) was included as a time-dependent variable. Continuous and categorical levels (<0.10, , , and > 1.0 infective bite/day) of the -week mean dose were analyzed. The cumulative dose of infective bites was derived using the duration of exposure variable and the corresponding daily mean entomologic inoculation rate estimates for each -week interval. Thus, for a child with a 8-day duration of exposure, the cumulative dose = (first -week mean daily entomologic inoculation rate X 14 days) + (second -week mean daily entomologic inoculation rate X 14 days). Age. Younger children, with less acquired immunity, generally experience higher density P. falciparum infections (6, 8). Age was derived using the child's birth date reported by the mother. Baseline parasitemia density. To indirectly control for unmeasured genetic, immunologic, and nutritional factors likely to influence the rate and density of parasitemia, each child's parasitemia density at enrollment was included in all analyses as a log-transformed continuous variable. Statistical analyses Two analytical approaches investigated the associations between the dose of inoculation and the incidence rate of first recurrent and first recurrent highdensity parasitemia. We first conducted an analysis consistent with an ecologic study, since all entomologic data were collected at the community level, not the individual level (34). Weighted least-squares linear regression, adjusted for each cohort's mean age, examined the amount of variation in cohort incidence rates explained by the mean cumulative dose for each cohort. Weights were calculated as the inverse of the variance (number of person-weeks /number of cases) of each cohort's incidence rate (35). In the second approach, the mean dose derived for each individual child was included as a time-dependent covariate in a Cox proportional hazards regression model. An analysis of third variables (age, baseline parasitemia density, sex, and village of residence) thought to potentially confound or modify the effect of the mean dose was also conducted. Multiple linear regression models assessed the influence of the cumulative dose, duration of exposure, and age on the density of first recurrent parasitemia. The dependent variable in all linear regression models was each child's log-transformed recurrent parasitemia density. Due to the study design, the duration of exposure was not truly continuous and was categorized as <14, 15-8, 9-4, 43-56, and days. Tests were conducted to compare the use of an indicator variable with a grouped linear dose-response variable for the duration of exposure. The likelihood ratio test validated the statistical appropriateness of the grouped Linear variable, thus permitting one term for the duration of exposure in all linear regression models. In the final models selected, variance inflation factors were obtained to assess multicollinearity, particularly between the cumulative dose and the duration of exposure. No significant multicollinearity was detected (no variance inflation factor > 3.0; each model's mean variance inflation factor <.5).

5 Plasmodium falciparum Transmission in Kenya 949 Cross-product terms were also included in the linear regression models to test whether the effect of the cumulative dose was dependent upon the levels of duration of exposure, age, or baseline parasitemia density. In models containing interaction terms, the main effects were centered by subtracting the mean values of cumulative dose, age, and baseline parasitemia density from each individual's value. Finally, tests for quadratic effects (cumulative dose and age ) were conducted separately in models with and without the interaction terms described above. RESULTS Description of the study sample At the respective enrollment times of the 0 cohorts, 958 eligible children were recruited into the study between February 1986 and July Baseline characteristics of each cohort (age, village of residence, sex, prevalence, and density of blood-stage P. falciparum infection) have been described (7). After exclusion of 96 children with positive day 14 blood films (n = 49) or who dropped out of the study prior to day 14 (n = 47), 86 individuals remained. Characteristics of the 0 cohorts are described in table 1. There were no significant differences in age, sex, or baseline parasitemia density between the 96 children excluded up to day 14 and the 86 children who remained for follow-up. In addition, there were no significant differences between the composition of the study sample at day 0 and day 14 (table 1). Distribution of dose and duration of exposure The cumulative dose and duration of exposure to sporozoite inoculation experienced by each cohort are presented in table. During follow-up of the 86 children, the mean cumulative dose was 3 (3) inoculations. An extended period of highest mean dose (entomologic inoculation rate > 1.0 infective bite/day) occurred between April and July 1986, while an extended period of lowest mean dose (entomologic inoculation rate < 0.15 infective bite/day) occurred between August and October 1986 (figure 1) (8). During the course of the study, the maximum daily mean dose was five infective bites/day for up to a -week duration (cohorts 4-7) and as low as zero infective bites/day for up to 4 weeks (cohorts 3 and 9-11). TABLE 1. Description of the 0 cohorts of children originally enrolled, cleared of their malaria parasites with drug treatment (day 0), and subsequently followed from a parasitemia-free time point (day 14) through the next 70 days during a Plasmodium falciparum incidence study in western Kenya, Cohort Enrollment month February March April May June July July August September October November December January February March April May June July August No. at dayo* No. at day 14t Mean age (years)t 3.0(1.6) 3.3(1.5).6(1.7) 3.0(1.6) 3.0(1.5).5(1.4) 3.(1.6) 3.1 (1.8).7(1.6).4(1.5).9(1.8) 3.3(1.5) 3.4(1.4).6(1.3).4(1.6) 3.0(1.7).6(1.4) 3.0(1.5).8(1.7) 3.(1.7) % male Overall (1.6) * Number of children originally enrolled into each cohort and present at day 0. t Number of children in each cohort with a negative day 14 blood film and thus atriskforrecurrent parasitemia. t Mean age of children in each cohort with a negative day 14 blood film. Numbers in parentheses, standard deviation. 5

6 9 McElroy et al. TABLE. Distribution of cumulative dose of sporozoite inoculation and mean duration of exposure (DOE) experienced by the children of 0 separate cohorts prior to their first recurrent Plasmodium falciparum parasitemia in western Kenya, Overall Mean Cumulative dose* Median Minimum Maximum (15)* 34(11) 7 (13) 19 (10) 0(9) 39 (0) 39 (3) 43() 43(17) 43(18) 36(18) 3 (19) 45(17) 51 (19) 33(17) 8 (15) 7 (14) 4(11) 37 (16) 4 (19) 36(19) * Cumulative number of inoculations experienced by cohort members during the exposure window (day 0 through 14 days prior to first detectable P. falciparum parasitemia). f Mean number of exposure days experienced by cohort members between day 0 and the date 14 days prior to first detectable recurrent P. falciparum parasitemia. X Numbers in parentheses, standard deviation. ate D 'o s 0 CO g o5 c o u s c -wks 1-r Inc. rate = (mean) SE(0,) = 0.10 I Mean infective bites per cohort B Aggregate analysis of infective bite exposure Results of the aggregate analysis are presented in figure 3. Weighted least-squares linear regression analysis of the individual study cohorts demonstrated a significant linear association between each cohort's mean cumulative dose exposure and the incidence rate of first recurrent parasitemia (r = 0.359, p = 0.007; figure 3A). The amount of variation in the incidence rate explained by the mean cumulative dose did not increase after the inclusion of the mean age of each cohort as a covariate in the model (r = 0.380, p = 0.0). In this regression analysis, every 10-unit increase in the mean cumulative dose corresponds to an increase of 3.3 cases per 100 person-weeks. The strength of the linear relation was stronger for the incidence of higher density parasitemia (r = 0.546, p < 0.001; figure 3B). The addition of each cohort's mean age to this regression model decreased the amount of unexplained variation in the incidence rate (r = 0.716, p < 0.001). In this model, a 10-unit increase in the mean cumulative dose among cohorts with a mean age of and 4 years corresponds to an O Inc. rate = (mean) SE (p,) = Mean infective bites per cohort FIGURE 3. Weighted least-squares linear regression analysis of the relation between the mean sporozoite inoculation exposure experienced by each separate cohort and the incidence rates of first recurrent parasitemia (A) and first recurrent high-density (>5,000//il) parasitemia (B) among Kenyan children in Inc. rate, incidence rate; SE, standard error; wks, weeks. increase of.4 and. cases per 100 person-weeks, respectively. Effect of covariates at the individual level The overall 70-day cumulative incidence of first recurrent P. falciparum parasitemia among the 86 children was 88.5 percent (95 percent confidence interval (CI) ). The estimated incidence rate was 17. cases per 100 person-weeks at risk. Age as a

7 Plasmodium falciparum Transmission in Kenya 951 categorical variable was not associated with the rate of recurrent parasitemia (table 3) nor was age significant when included as a continuous variable (data not presented). However, increased parasitemia density at the time of enrollment was associated with a slightly increased rate of recurrent parasitemia. Compared with children from the village of Got Onyango, children from Abwao had an elevated rate of recurrent parasitemia (rate ratio (RR) = 1.40, 95 percent CI ), while residence in Saradidi (RR = 0.73, 95 percent CI ) and Luoro (RR = 0.77, 95 percent CI ) was associated with reduced rates. Males experienced a slightly higher rate of recurrent infection. These rate ratio estimates were adjusted for the time-dependent dose of infective mosquito bites. The overall 70-day cumulative incidence of first recurrent high-density P. falciparum (^5,000//A1) among the 86 children was.5 percent (95 percent CI ). The estimated incidence rate was 4.4 cases per 100 person-weeks at risk. In these analyses (table 3), age was significantly associated with the rate of high-density parasitemia, with children 18-3 months of age at highest risk for the outcome relative to the oldest children (RR =.88, 95 percent CI ). Baseline parasitemia density and village were again associated with a higher rate of parasitemia S: 5,000//A1. Compared with Got Onyango, Abwao and Rariw both had increased rates (RR =.13, 95 percent CI (Abwao); RR = 1.67, 95 percent CI (Rariw)). Sex was not associated with the incidence rate of high-density parasitemia. These rate ratio estimates were also adjusted for the time-dependent dose of infective mosquito bites. Infective bite exposure at the individual level Analysis of the mean dose as both a continuous and categorical time-dependent exposure variable detected significant associations with the incidence rates of recurrent parasitemia (table 4). Each one-unit increase in the mean dose was associated with a 4 percent higher rate of recurrent parasitemia (RR = 1.4) and a 6 percent higher rate of recurrent high-density parasitemia (RR = 1.6) after adjustment for other covariates. At the categorical level, among children with a mean dose of ^1.0 infective bite/day, the rate of recurrent parasitemia was.-fold higher in comparison with children who had a mean dose of <0.10 infective bite/day. Likewise, the rate of recurrent highdensity parasitemia was.8-fold greater among children with the highest mean dose exposure level as compared with the referent. TABLE 3. Adjusted rate ratios for first recurrent and first recurrent high density ( 5,0Ou/jJ) parasitemia among 86 children in western Kenya, Variable No. Infection (all levels) RRM 95% CI* Infection (a5,000/nl) RRt 95% CI BPD* Age (months) Sex Female Male Village Got Onyango Miyare Saradidi Abwao Rariw Luoro * RR, rate ratio; CI, confidence interval; BPD, baseline parasitemia density (log-transformed values). t Adjusted for other covanates in the table plus infective bite dose (a time-dependent, continuous variable).

8 95 McElroy et al. TABLE 4. Results of multivariate Cox proportional hazards regression examining associations between mean dose of infective bite exposure and all levels and high density levels (>5,000/uJ) of recurrent Plasmodium falciparum blood stage infection among 86 children in western Kenya, * Mean doset Continuous Ordered < : Mo ±, RR Infection (all levels) 95% Cl RR Infection (a5,000/nl) 95% Cl * Adjusted for baseline parasitemia density, age (continuous), sex, and village of residence. t Mean dose of infective bite exposure per day experienced by a child during follow-up. i Number of -week exposure periods contributed by the 86 children included in the follow-up period. RR, rate ratio: Cl, confidence interval. Predictors of parasitemia density Multiple linear regression analyses were conducted to study the relations between parasitemia density and cumulative dose while accounting for other important covariates. Results of a linear model without interaction terms are presented in table 5. Cumulative dose was positively associated with density, while duration of exposure and age were both inversely associated with parasite density. The village of residence, namely, Abwao and Rariw, and baseline parasitemia density were marginally associated with density and were included in all regression models tested. Sex was not a significant predictor of parasite density in any model and was omitted. To test for suspected interactions, we individually analyzed three cross-product terms (cumulative dose X duration of exposure, cumulative dose X age, and cumulative dose X baseline parasitemia density) in the linear regression model presented in table 5. The cumulative dose X duration of exposure term was significant when added to this model (/3 = 0.370, p < 0.001), but the cumulative dose X age (J3 = 0.058, p = 0.6) and cumulative dose X baseline parasitemia density (/3 = 0.048, p = 0.539) terms were not significant. However, we suspected that the association between cumulative dose and parasitemia density was modified by the effect of duration of exposure if age was taken into account. Thus, a threeway interaction term (cumulative dose X duration of exposure X age) was included (table 6). The effect of cumulative dose was inversely related to parasite density as the duration of exposure increased (j3 = , p < 0.001) and as age increased (/3 = -0.9, p = 0.00). The effect of a one-unit increase in the cumulative dose in predicting higher parasitemia density was less dramatic when the duration of exposure was more prolonged, particularly among older children. TABLE 5. Results of multiple linear regression model of log-transformed parasitemia density values regressed on cumulative dose, duration of exposure, age, baseline parasitemia density, and village of residence among 86 Kenyan children, Variable (units for coefficient) Intercept Cumulative dose (log no.) Duration of exposure (days)t Age (years) Baseline parasitemia density (log no.) Village Miyare Saradidi Abwao Rariw Luoro P % Cl* to to to to to to to to to to p value < * Cl, confidence interval. t Coded as follows: 1-14 days = 0,15-8 days = 1, 9-4 days =, days = 3, and days = 4. X Log-transformed values of the number of asexual parasites/u.l of blood for each child at the time of enrollment into the study. Total ft 0.083

9 Plasmodium falciparum Transmission in Kenya 953 TABLE 6. Results of multiple linear regression model of log-transformed parasitemia density with interaction terms for cumulative dose, duration of exposure, and age among 86 Kenyan children, Variable (units for coefficient) Intercept Cumulative dose (log no.)t Duration of exposure (days)}: Age (years)t Baseline parasitemia density (log no.) Village Miyare Saradidi Abwao Rariw Luoro Cumulative dose x duration of exposure Cumulative dose x age Duration of exposure x age Cumulative dose x duration of exposure xage P % Cl*.175 to to to to to to to to to to to to to to p value < Total ft 0.13 Cl, confidence interval. t Cumulative dose and age were centered by subtracting overall mean values of and 3.0, respectively. % Coded as follows: 1-14 days = 0, 15-8 days = 1, 9-4 days =, days = 3, and days = 4. Log-transformed values of the number of asexual parasites/pi of blood for each child at the time of enrollment into the study. Tests for quadratic effects (cumulative dose and age ) were also conducted. Neither quadratic term was significant in any main effects model nor did the addition of either quadratic term alter the estimated values of coefficients in the table 6 model. For example, the addition of cumulative dose resulted in only the slightest change to the estimated coefficient of the cumulative dose X duration of exposure X age interaction term (/3 = 0.101, p = 0.007). Likewise, the addition of age resulted in no appreciable change to the estimated coefficient of the cumulative dose X duration of exposure X age term (/3 = 0.068, p = 0.076). DISCUSSION We have explored the relation between the frequency of exposure to P. falciparum sporozoite inoculation and the subsequent incidence and density of parasitemia in young Kenyan children. An improved understanding of the forces that influence the dynamics of the preerythrocytic and erythrocytic stages of P. falciparum infection may facilitate the development of effective control strategies. Results of these analyses offer limited yet provocative insight into a contentious issue in malariology: whether reductions in transmission intensity, without total elimination, will result in less morbidity and mortality. Using each cohort as a unit of analysis, we found that approximately 36 percent (r = 0.359; figure 3A) of the variation in P. falciparum incidence rates was explained by the mean cumulative dose of infective bites. The addition of each cohort's mean age as an explanatory variable to this regression model did not reduce the amount of unexplained variation in incidence rates. This provides additional support for our conclusion that age is not a significant risk factor for the rate of P. falciparum infection among children in western Kenya (table 3) (8). However, analysis of recurrent high-density parasitemia (^5,000/JUJ) showed that the mean cumulative dose accounted for approximately 55 percent (r = 0.546) of the variation in such incidence rates. In this regression model, the amount of explained variation in the incidence rates increased to 7 percent (r = 0.716) upon the addition of the mean age of the cohort as an independent variable. At the individual level, two measures of infective bite dose (mean dose and cumulative dose) were found to be significantly associated with the rates of recurrent parasitemia and density of recurrent parasitemia. Each one-unit increase of the mean dose was associated with a 4 percent and a 6 percent higher rate of all-level and high-density level parasitemia, respectively (table 4). The associations detected between the infective bite dose and the incidence rate were not confounded by age, baseline parasitemia density, village of residence, or sex. The observed differences in the time to recurrent parasitemia, assuming that sporo-

10 954 McElroy et at. zoite exposure did in fact occur at the lower doses of inoculation, may be due to children's ability to cope with smaller numbers of preerythrocytic stage parasites via acquired immunologic mechanisms. Results of multiple linear regression analyses provided the most compelling evidence of the importance of cumulative dose and duration of exposure in predicting parasitemia density, particularly among the very youngest children (tables 5 and 6). Prior to the addition of interaction terms (table 5), the most significant predictors of parasitemia density were cumulative dose {p < 0.001), duration of exposure (p < 0.001), and age (p = 0.006). While older age was also associated with lower parasitemia density, a significant interaction between cumulative dose and age was not detected until duration of exposure was also included in a three-way interaction term. Thus, even though older children generally experienced lower density recurrent parasitemias, they responded to the effect of an increased unit of cumulative dose that was dependent upon the duration of time (duration of exposure) over which the infective bites were administered. For example, a 1-year-old child who experienced 30 infective bites during 14 days will have a first recurrent parasitemia density over 10 times greater than another 1-year-old child who received the same number of infective bites delivered over 4 days (figure 4). This age-dependent interaction detected between cumulative dose and duration of exposure (p = 0.011) has not been previously reported. After controlling for transmission intensity, all children in this study became reinfected by P. falciparum at similar rates, regardless of their degree of acquired immunity as indirectly measured by age (table 3, alllevel infections). However, age was associated with the rate of recurrent high-density parasitemia. Children between 18 and 3 months of age experienced the highest rate of recurrent high-density parasitemia (S:5,000//^l) as compared with the referent group (RR =.88, 95 percent CI ). These results are consistent with the authors' earlier findings of no association between age and the prevalence of P. falciparum parasitemia upon enrollment but a clear inverse association between age and parasitemia density (7). Recent observational studies from several areas in Africa suggest a non-uniform level of risk for severe malarial anemia and cerebral malaria among children 6 months to 6 years of age (). Those observations, in conjunction with our current results, suggest that acquired immunologic mechanisms do not inhibit the rate of recurrent P. falciparum parasitemia, even after 5 years of intense exposure (table 3). Rather, immunologic mechanisms improve with age and limit the density of parasitemia, thereby placing an older child Cumulative no. of Infective Anopheles exposures Cumulative no. of infective Anopheles exposures 1 yr FIGURE 4. Dose- and time-dependent relations between infective Anopheles inoculation and Plasmodium falciparum parasitemia among children of three different ages from the same village (Got Onyango) of western Kenya, A, 14-day exposure; B, 4-day exposure, yr, year. (4-6 years of age) at a reduced risk for severe malarial anemia associated with mass hemolysis. This acquired immunity may be a consequence of cumulative exposure to P. falciparum over time or of an age-dependent process independent of exposure history (36). The associations between inoculation frequency and parasite density detected in this study may be biased toward the null. It is unlikely that the entomologic exposure data obtained at the community level are entirely representative of all individual-level exposures. However, in applying the ecologic data to individual children, any misclassification of the inoculation exposure was likely nondifferential and thus resulted in an underestimate of the true strength of association. Possible misclassification of the entomologic exposure data is suggested by the nonuniformity yr

11 Plasmodium falciparum Transmission in Kenya 955 of the estimated rate ratios among the six villages (table 3). The stronger associations between particular villages and the incidence of both outcomes, after adjusting for infective dose, are probably due to an inability to capture the finer differences in transmission intensity when entomologic data are collected at the aggregate level. Further investigation into the role of transmission intensity and malarial outcomes will be improved using sporozoite inoculation data obtained from individual households, thereby reducing exposure misclassification. Colleagues are currently conducting such studies in western Kenya and Mali. The relations between controlled doses of sporozoite inoculation and various malarial outcomes have been previously investigated. Data in those studies were obtained from experimentally induced infections (i.e., challenge with laboratory-reared infective mosquitoes or via injection of blood-stage parasites) in nonimmune adults (17-0). In addition, a homologous strain of parasite was always used. Most recently, Glynn et al. (37) have used malaria therapy data obtained over several decades from neurosyphilis patients to explore whether sporozoite dose-dependent outcomes exist. In one analysis, no associations were found between the infective dose and any malaria outcome, including the time until the first detectable parasitemia. Another analysis detected an association only between the dose and the time until parasitemia, but the parasite species was Plasmodium vivax, not P. falciparum (38). The inconsistencies between the analyses of Glynn and colleagues and our findings are not surprising given these dramatic differences in the predictor variables. Finally, our results offer additional parameters for including in a systems analysis approach to modeling patterns of mild and severe malaria incidence as elaborated by Gupta et al. (39). A systems approach to understanding the relations among all factors believed to influence the dynamics of P. falciparum parasitemia, including transmission intensity and parasite virulence, will offer improved opportunities to develop and evaluate the effects of new antimalarial strategies (40). In addition, this approach will assist laboratory scientists' thinking regarding the nature of protective immune responses against severe malarial anemia and cerebral malaria. Further investigation of these relations will be facilitated using well-defined clinical malaria endpoints collected from multiple geographic locations of varying infective bite intensity. ACKNOWLEDGMENTS Supported by the US Naval Medical Research and Development Command work unit 6.3A HFX.1433 and the US Army Medical Research and Development Command. John C. Beier received partial support from NTH grant AI In conducting the research described in this report, protocols involving human subjects were approved by ethical committees of the Walter Reed Army Institute of Research and the Kenya Medical Research Institute. REFERENCES 1. World Health Organization. Control of tropical diseases: malaria control. Geneva: World Health Organization Office of Information, Marsh K. Malaria a neglected disease? Parasitology 199; 104:S Miller LH, Good MF, Milon G. Malaria pathogenesis. Science 1994;64: Grau GE, de Kossodo S. Cerebral malaria: mediators, mechanical obstruction, or more? Parasitol Today 1994; 10: Field JW, Niven JC. A note on prognosis in relation to parasite counts in acute subtertian malaria. Trans R Soc Trop Med Hyg 1937;30: Phillips RE. The importance of anaemia in cerebral and uncomplicated falciparum malaria: role of complications, dyserythropoiesis, and iron sequestration. Q J Med 1986;58: Warrel DA, Molyneux ME, Beales PF. Severe and complicated malaria. Trans R Soc Trop Med Hyg 1990;84:l Berendt AR, Turner GDH, Newbold CI. Cerebral malaria: the sequestration hypothesis. Parasitol Today 1994;10: Hoffman SL, Rustama D, Punjabi NH, et al. High-dose dexamethasone in quinine-treated patients with cerebral malaria: a double-blind, placebo-controlled trial. J Infect Dis 1988; 158: Miller LH. Genetically determined human resistance factors. In: Wernsdorfer WH, McGregor IA, eds. Malaria: principles and practice of malariology. London: Churchill Livingstone, 1988: McGregor IA, Wilson RJM. Specific immunity: acquired in man. In: Wernsdorfer WH, McGregor IA, eds. Malaria: principles and practice of malariology. London: Churchill Livingstone, 1988: Glynn JR, Lines J, Bradley DJ. Impregnated bednets and the dose-severity relationship in malaria. Parasitol Today 1994; 10: Snow RW, Marsh K. Will reducing Plasmodium falciparum transmission alter mortality among African children? Parasitol Today 1995;11: Alonso PL, Lindsay SW, Armstrong JRM, et al. A malaria control trial using insecticide-treated bed nets and targeted chemoprophylaxis in a rural area of The Gambia, West Africa. 6. The impact of the interventions on morbidity and mortality. Trans R Soc Trop Med Hyg 1993;87: D'Alessandro U, Olaleye BO, McGuire W, et al. Mortality and morbidity from malaria in Gambian children after introduction of an impregnated bednet programme. Lancet 1995; 345: Beach RF, Reubush TK, Sexton JD, et al. Effectiveness of permethrin-impregnated bednets and curtains for malaria control in a holoendemic area of western Kenya. Am J Trop Med Hyg 1993;49: Boyd MF, Kitchen SF. Observations on induced falciparum malaria. Am J Trop Med 1937;17: Coatney GR, Cooper WC, Young MD, et al. Studies on human malaria. I. The protective action of sulfadiazine and sulfapyrazine against sporozoite-induced falciparum malaria. Am J Hyg 1947;46:

12 956 McElroy et al. 19. Covell G, Nicol WD, Shute PG, et al. Studies on a West African strain of Plasmodium falciparum. U. The efficacy of paludrine (proguanii) as a therapeutic agent. Trans R Soc Trop Med Hyg 1949;4: Rickman LS, Jones TR, Long GW, et al. Plasmodium falciparum-infected Anopheles stephensi inconsistently transmit malaria to humans. Am J Trop Med Hyg 1990;43: Hoffman SL, Sacci JB. Rationale and approaches to constructing pre-erythrocytic malaria vaccines. In: Powell MF, Newman MI, eds. Vaccine design: the subunit and adjuvant approach. New York: Plenum Press, 1995: Beier JC, Perkins PV, Onyango FK, et al. Characterization of malaria transmission by Anopheles (Diptera:Culicidae) in western Kenya in preparation for malaria vaccine trials. J Med Entomol 1990;7: Bockarie MJ, Service MW, Barnish G, et al. Vectorial capacity and entomological inoculation rates of Anopheles gambiae in a high rainfall forested area of southern Sierra Leone. J Trop Med 1995 ;46: Snow RW, Bastos de Azevedo I, Lowe BS, et al. Severe childhood malaria in two areas of markedly different P. falciparum malaria transmission in East Africa. Acta Trop 1994; 57: Mbogo CNM, Snow RW, Khamala CPM, et al. Relationship between Plasmodium falciparum transmission by vector populations and the incidence of severe disease at nine sites on the Kenyan coast. Am J Trop Med Hyg 1995;5: McElroy PD, Beier JC, Oster CN, et al. Predicting outcome in malaria: correlation between rate of exposure to infected mosquitoes and level of Plasmodium falciparum parasitemia. Am J Trop Med Hyg 1994;51: Beier JC, Oster CN, Onyango FK, et al. Plasmodium falciparum incidence relative to entomologic inoculation rates at a site proposed for testing malaria vaccines in western Kenya. Am J Trop Med Hyg 1994;: Beadle C, McElroy PD, Beier JC, et al. Impact of transmission intensity and age on Plasmodium falciparum density and associated fever: implications for malaria vaccine trial design. J Infect Dis 1995;17: Beier MS, Schwartz IK, Beier JC, et al. Identification of malaria species by ELISA in sporozoite and oocyst infected Anopheles from western Kenya. Am J Trop Med Hyg 1988; 39: Greenwood BM, Armstrong JR. Comparison of two simple methods for determining malaria parasite density. Trans R Soc Trop Med Hyg 1985;85: Molineaux L, Muir DA, Spencer HC, et al. The epidemiology of malaria and its measurement. In: Wernsdorfer WH, McGregor LA, eds. Malaria: principles and practice of malariology. London: Churchill Livingstone, 1988: Trape JF, Peelman P, Morault-Peelman B. Criteria for diagnosing clinical among a semi-immune population exposed to intense and perennial transmission. Trans R Soc Trop Med Hyg 1985;79: Greenwood BM, Bradley AK, Greenwood AM, et al. Mortality and morbidity from malaria among children in a rural area of The Gambia, West Africa. Trans R Soc Trop Med Hyg 1987;85: Rothman KJ. Multivariate analysis. In: Modern epidemiology. Boston: Little, Brown and Co, 1986: Selvin S. Estimates of risk from follow-up data. In: Statistical analysis of epidemiologic data. New York: Oxford University Press, Inc, 1996: Baird JK, Jones TR, Danudirgo EW, et al. Age-dependent acquired immunity against Plasmodium falciparum in people having two years exposure to hyperendemic malaria. Am J Trop Med Hyg 1991;45: Glynn JR, Collins WE, Jeffery GM, et al. Infecting dose and severity of falciparum malaria. Trans R Soc Trop Med Hyg 1995;89: Glynn JR, Bradley DJ. Inoculum size, incubation period, and severity of malaria. Analysis of data from malaria therapy records. Parasitology 1995;110: Gupta S, Hill AVS, Kwiatkowski D, et al. Parasite virulence and disease patterns in Plasmodium falciparum malaria. Proc Natl Acad Sci USA 1994,91: Koopman JS. Emerging objectives and methods in epidemiology. Am J Public Health 1996;86:630-.