• View in gallery

    School sampling in Leogane Commune, Haiti, October 1999–June 2000 and March–May 2001. CDC = Centers for Disease Control and Prevention.

  • View in gallery

    Wuchereria bancrofti infection prevalence in Leogane Commune, Haiti by student home community (October 1999–June 2000 and March–May 2001).

  • 1

    Michael E, Bundy DAP, Grenfell BT, 1996. Re-assessing the global prevalence and distribution of lymphatic filariasis. Parasitology 112 :409–428.

    • Search Google Scholar
    • Export Citation
  • 2

    World Health Organization, 1997. World Health Assembly Resolution 50.29. Geneva: World Health Organization.

  • 3

    Ramzy RM, Hafez ON, Gad AM, Faris R, Harb M, Buck AA, Weil GJ, 1994. Efficient assessment of filariasis endemicity by screening for filarial antigenaemia in a sentinel population. Trans R Soc Trop Med Hyg 88 :41–44.

    • Search Google Scholar
    • Export Citation
  • 4

    World Health Organization, 1998. Research of Rapid Geographic Assessment of Bancroftian Filariasis. Geneva: World Health Organization. TDR/TDF/COMDT/98.2).

  • 5

    World Health Organization, 1999. Report of a WHO Informal Consulation on Epidemiologic Approaches to Lymphatic Filariasis Elimination: Initial Assessment, Monitoring and Certification. Geneva: World Health Organization. WHO/FIL/99.195.

  • 6

    World Health Organization, 1999. Guidelines for Certifying Lymphatic Filariasis Elimination (Including Discussion of Critical Issues and Rationale). Geneva; World Health Organization. WHO/FIL/99.197.

  • 7

    Breslow NE, Clayton DG, 1993. Approximate inference in generalized linear mixed models. J Am Stat Assoc 88 :9–25.

  • 8

    Agresti A, Booth JG, Hobert JP, Caffo B, 2000. Random-effects modeling of categorical response data. Sociol Methodol 30 :27–80.

  • 9

    Littell RC, Milliken GA, Stroup WW, Wolfinger RD, 1996. SAS® System for Mixed Models. Cary, NC: SAS Institute, Inc., 423–460, 491–495.

  • 10

    Onapa AW, Simonesen PE, Pedersen EM, Okello DO, 2001. Lymphatic filariasis in Uganda: baseline investigations in Lira, Soroti and Katakwi districts. Trans R Soc Trop Med Hyg 95 :161–167.

    • Search Google Scholar
    • Export Citation
  • 11

    Sherchand JB, Obsomer V, Thakur GD, Hommel M, 2003. Mapping of lymphatic filariasis in Nepal. Filaria J 2 :7. (http://www.filariajournal.com/content/2/1/7).

    • Search Google Scholar
    • Export Citation
  • 12

    Molyneux DH, Zagaria N, 2002. Lymphatic filariasis elimination: progress in global programme development. Ann Trop Med Parasitol 96 :S15–S40.

    • Search Google Scholar
    • Export Citation
  • 13

    Evans DB, Gelband H, Vlassoff C, 1993. Social and economic factors and the control of lymphatic filariasis: a review. Acta Trop 53 :1–26.

    • Search Google Scholar
    • Export Citation
  • 14

    Hunter JM, 1992. Elephantiasis: a disease of development in northeast Ghana. Soc Sci Med 35 :627–649.

 

 

 

 

COMMUNITY- AND INDIVIDUAL-LEVEL DETERMINANTS OF WUCHERERIA BANCROFTI INFECTION IN LEOGANE COMMUNE, HAITI

View More View Less
  • 1 Department of Epidemiology and Department of Biostatistics, Rollins School of Public Health, Emory University, Atlanta, Georgia; Division of Parasitic Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia; Hôpital Sainte Croix, Leogane, Haiti

We conducted a school-based assessment of the geographic distribution of Wuchereria bancrofti infection in Leogane Commune, Haiti, using the immunochromatographic test. In multivariate analyses performed using generalized linear mixed models, children attending schools in the foothills and plains were 3.95 (95% confidence interval [CI] = 1.28–12.23) and 23.56 (95% CI = 8.99–61.79) times as likely to be infected, respectively, as children attending mountain schools. Infection prevalence decreased with increasing altitude, but some local foci of infection were detected at higher altitudes. Higher school tuition, a marker of socioeconomic status (SES), was not associated with decreased infection prevalence. Our results indicate that although the force of infection in Leogane Commune is greatest below 70 meters above sea level, higher altitude communities are not exempt from infection. Lymphatic filariasis (LF) elimination programs should consider extending infection mapping activities to presumed non-LF altitudes. In addition, higher SES does not confer protection against W. bancrofti infection.

INTRODUCTION

Lymphatic filariasis (LF), a mosquito-transmitted disease caused by the parasitic worms Wuchereria bancrofti, Brugia malayi, and Brugia timori, affects an estimated 120 million persons throughout the tropics.1 In 1997, the World Health Assembly passed a resolution calling for the “global elimination of lymphatic filariasis as a public health problem,”2 prompting the establishment of LF elimination programs in numerous countries.

The geographic distribution of filarial infection is nonuniform.3,4 Since the factors influencing this distribution have not been identified, elimination programs initially conduct nationwide blood surveys to construct national LF prevalence maps.5 However, certification of elimination requires intensive screening for infection in areas at high risk for LF transmission, and the cost of refining the initial national surveys to precisely demarcate at-risk areas and identify regions for certification sampling may be prohibitive.6 Consequently, there is a need to ascertain which environmental, geographic, and demographic factors determine the distribution of filarial infection and develop a tool for prediction of community risk of filarial infection using easily-obtainable information.

In 1999, the Centers for Disease Control and Prevention (CDC) initiated an LF elimination demonstration program in Leogane Commune, Haiti, in collaboration with Hôpital Sainte Croix, the main health facility in the Commune, and Notre Dame University (Notre Dame, IN). Leogane Commune is located on Haiti’s southern peninsula and covers an area of approximately 400 km2. Home to 150,000–200,000 people, the Commune is divided into 13 sections (Hôpital Sainte Croix, unpublished data). The three sections in the coastal plain (1, 2, and 3) are the most populous; the remaining 10 sections are mountainous and more sparsely populated. Initial program activities included an assessment of the geographic distribution of W. bancrofti infection in the Commune. We describe this assessment and evaluate the relationships between infection and potential geographic, socioeconomic, and demographic risk factors.

METHODS

Study population and sampling methodology.

We conducted a cluster survey to assess the baseline geographic distribution of W. bancrofti infection in Leogane Commune using primary schools as the sampling unit. We assumed that the infection prevalence of a school reflected the prevalence of infection in the surrounding community, since school children lived near their schools or, in the case of some mountain schools, in nearby communities resembling those in which their schools were located. Schools were selected to yield good geographic coverage of the Commune. Our primary grid sampling strategy was supplemented by oversampling schools in Leogane town (the principal urban area in the Commune), its immediate surroundings and potential program monitoring sites, and by convenience sampling of schools in remote mountain areas where finding schools became difficult. Children between the ages of 5 and 11 years were eligible to participate in the survey. At each selected school, we assessed infection status in all eligible children who presented themselves with a parent or guardian on a specified test day. Prior to testing, children and their parents/guardians were educated about the LF elimination program and the purpose of the school prevalence survey, and informed consent was obtained.

The survey protocol and consent form were reviewed and approved by both the CDC and Hôpital Sainte Croix institutional review boards (IRBs). The IRB of Emory University reviewed and approved the proposed analyses as secondary analyses of pre-existing data.

Grid sampling.

Twenty-five schools were selected using a grid sampling scheme. Using ArcView software version 3.2 (Environmental Systems Research Institute, Inc., Redlands, CA), a grid of rectangles of equal size was overlaid on a map of the Commune (Figure 1), excluding remote mountain areas. A point within each grid rectangle was then chosen at random and its longitude and latitude were read from the map. Using a handheld global positioning system (GPS) unit (Magellan unit; Thales Navigation, Inc., Santa Clara, CA) to find each random point in the field, an advance team located the primary school(s) nearest that point (Figure 1). We aimed to enroll at least 100 children per grid rectangle. If a single school with 100 or more eligible children did not exist near the random point, we selected multiple schools whose combined eligible children numbered more than 100. On occasion, in sparsely populated areas with few schools, enrollment fell short of 100 students. Once schools were identified, the team met with each headmaster to discuss the purpose of the school prevalence survey and encourage him to allow his school to participate. No school declined to participate.

Monitoring site schools.

Three communities under consideration as program monitoring sites fell under the grid but had not contributed schools to the survey. Since monitoring sites were selected, in part, based on community prevalence of infection, all six schools in these communities were surveyed (Figure 1).

Mountain school selection.

Many mountainous areas were remote and rugged, and could only be accessed on foot. The grid sampling strategy could not be executed in these areas because locating random points and nearby schools on foot proved unfeasible. To obtain good coverage of the mountainous portion of the Commune, the advance team met with community health workers (CHWs) from sections 6–10 and 13–15 to enlist their help in identifying schools in their sections. All 17 schools identified by the CHWs present at the meeting participated in the prevalence survey (Figure 1).

Leogane town and immediate surroundings.

Prior to establishing the LF elimination program in Leogane Commune, CDC conducted a school-based randomized control trial to examine the efficacy of diethylcarbamazine (DEC) and albendazole in treating W. bancrofti and intestinal helminth infections in 5–11-year-old children living in and around the town of Leogane (Beach M, unpublished data). Ten schools, five from the town of Leogane proper and five from surrounding communities (within 5 km of town), participated in the trial; 1,329 children were enrolled. Of these, 291 children from nine schools were evaluated for W. bancrofti infection at baseline using the same test used in our prevalence survey. Their test results were used as the school prevalence data for the town of Leogane and neighboring communities (Figure 1).

Antigen testing and individual-level data collection.

Each child’s birth date (if known), age, sex, and home community were recorded during the consent process before infection status assessment. W. bancrofti infection status was evaluated using an immunochromatographic test (ICT) (AMRAD, Sydney, New South Wales, Australia), a qualitative test for the presence of circulating W. bancrofti adult worm antigen. Capillary (fingerstick) blood (100 μL) was collected from each child and applied to an ICT card. The test result was read as antigen positive (infected) or antigen negative (uninfected) after 15 minutes. Infected children were treated with a single tablet (50 mg) of DEC and all children were treated with albendazole.

School-level data collection.

The longitude and latitude of each school were recorded using GPS (TSC1 Asset Surveyor with Pro XRS receiver and Pathfinder Office software, Trimble Navigation, Ltd., Sunnyvale, CA; and a Garmin eMap unit with Map Source software, Garmin International, Inc., Olathe, KS). Using these GPS-derived school coordinates, we read the altitude of each school in meters above sea level from a detailed topographic map of the Commune (scale = 1:50,000). We also used the geographic coordinates to categorize the schools by administrative section and topographic zone; in the latter case, schools were characterized as being located on the plains, in the foothills, or in the mountains based on school location and knowledge of the surrounding terrain.

Each headmaster was asked to provide information about 1) the school’s tuition, 2) the in-school nutrition program for students (if such a program existed), and 3) the communities served by the school.

Analytic methods.

All analyses were performed using SAS version 8.2 (SAS Institute, Inc., Cary, NC) and all maps were constructed using ArcView. Prior to modeling, descriptive analyses were performed on the independent and dependent (infection status) variables. Logistic generalized linear mixed modeling was conducted using SAS’s GLIMMIX macro.7–9 All models included random school effects to account for within-school correlation (correlation between students attending the same school). We first fit single-variable models to identify possible associations between infection status and individual independent variables.

All independent variables were included in the initial multivariate model. However, topographic zone, section, and altitude were highly collinear; including two (or all three) of these variables in the same model caused parameter estimate variances to become large and rendered the model unstable. As a result, although each variable provided important, different, and complimentary information, only one of the three could be included at a time in subsequent multivariate models.

We considered four multivariate mixed models containing topographic zone, section, quartiles of altitude, or alternate altitude ranges, respectively. All four models contained age, sex, tuition, nutrition program and random schools effects; all models also initially contained an age × sex interaction term. The age × sex interaction was non-significant in all models (all P > 0.70) and was therefore dropped, leaving only main effects models of the form

logit(πij)=β0β1ageij+β2genderij+β3tuitio+β4nutritionij+Σkβ(k+4)geographyijk+ρi

where j = 1, . , ni, with ni being the number of children in the ith school; i = 1, . , 57; πij denotes the probability of infection for the jth child from the ith school; geography denotes topographic zone (k = 1,2), section (k = 1,2 . 8), or altitude (k = 1,2,3 or 1,2,3,4); and ρi is the random effect of the ith school. The level of significance (α) was set at 0.05 for all multivariate models. The results from these mixed models should be interpreted given the random effects in the model; in the interests of brevity, this condition is implied rather than explicitly stated in the sections that follow.

RESULTS

Student characteristics.

We tested 3,318 students from 57 schools for W. bancrofti infection. The number of students tested per school ranged from 7 to 167 (median = 47, mean = 58). As shown in Table 1, the distribution of ages was relatively uniform, with slightly fewer younger children (five and six year-olds). Forty-nine percent of the children were male. The proportions of males and females in each age group were approximately equal.

School characteristics.

School tuition ranged from 40 to 600 gourdes/year, with a median of 182.5 gourdes/year (Table 1; ~ U.S. $2–35/year, median ~ U.S. $11/year). One-third of the schools had in-school nutrition programs for their students. Twenty-two (38.6%) of the schools were located in the mountains, 13 (22.8%) were foothills schools, and 22 (38.6%) were plains schools. Almost one-third of the schools were located ≤ 25 meters above sea level; more than half were located ≤ 75 meters above sea level. Thirteen (22.8%) schools were located > 400 meters above sea level.

The distributions of age and sex were similar in all three topographic zones. In contrast, the distribution of tuition varied by topographic zone. Plains schools accounted for 70% of schools in the highest quartile of tuition; 9 of 22 plains schools (40.9%), compared with 2 of 21 mountain schools (9.5%) and 2 of 13 foothills schools (15.4%), were in the highest tuition quartile. Of the 14 schools in the lowest quartile of tuition, seven were mountain schools and six were plains schools.

The distribution of nutrition programs also varied by tuition and topographic zone. Of 40 schools in the middle and highest quartiles of tuition, 30 (75%) did not have in-school nutrition programs. In contrast, more than 70% (10 of 14) of lowest quartile schools had such programs. Forty-eight percent (10 of 21) of plains schools had in-school nutrition programs, whereas only 16.7% (2 of 12) of foothills schools and 31.8% (7 of 22) of mountain schools had nutrition programs.

Distribution of W. bancrofti infection and single-variable model results.

Five hundred thirty-three children (16.1%) tested positive for W. bancrofti infection. W. bancrofti infection prevalence by student home community is shown in Figure 2. Table 2 summarizes the distribution of infection with respect to student and school characteristics and presents the single-variable model results. We observed modest associations between age, sex, and infection status. Children attending schools in the middle quartiles of tuition were less likely to be infected than children attending lowest quartile schools, whereas children attending highest quartile schools were almost three times as likely to be infected as children from lowest quartile schools. Seventy-nine percent (421 of 533) of infected children attended plains schools; children attending schools in the foothills and plains were 4 and 29 times as likely to be infected, respectively, as their counterparts in the mountains. Sections 1 through 3, which encompass the entire plains portion of the Commune as well as some foothills areas, had a combined infection prevalence of 30.9%. These three sections accounted for 92.2% of infections. In contrast, predominantly mountainous sections 6–15 had a combined prevalence of 2.4%. Mountain sections 7 and 14 and mountain/foothills section 15 accounted for 78.6% of the infections in this group. Children attending schools in sections 1, 2, 3, 7, 14, and 15 were more likely to be infected than children attending schools in sections 10 and 13, two of the most remote and rugged sections in the Commune (prevalence odds ratio [POR] range = 3.21–69.69). In contrast, children attending schools in sections 6, 8, 9, 11, and 12 were no more likely to be infected than children in the reference group. The prevalence of infection decreased dramatically with increasing altitude; we observed similar patterns with quartiles of altitude and absolute altitude levels.

Multivariate results.

Topographic zone remained strongly and significantly associated with infection status in the multivariate analyses (Table 3). Children attending schools in the foothills were almost 4 times as likely to be infected as their peers attending mountain schools, while children attending plains schools were more than 23 times as likely to be infected.

We observed very strong associations between infection status and the section in which a child’s school was located (Table 3). Children attending schools in sections 1, 2, 3, or 7 were substantially and significantly more likely to be infected than children attending schools in sections 10 and 13 (POR range = 26.82–62.06). In contrast, children attending schools in sections 6, 8, 9, 11, and 12 were as likely as children attending schools in the reference sections to be infected. Children attending schools in sections 14 and 15 were more than three times as likely as the reference children to be infected; however, neither association achieved statistical significance.

The association of infection with altitude also remained unattenuated in the multivariate analyses, for both quartiles of altitude and absolute altitude levels (Table 3).

Age continued to be associated with infection status in the multivariate analyses, regardless of the geographic variable included in the model. Seven and eight-year-old and 9–11-year-old children were 30–31% (P value range = 0.039–0.048) and 25–28% (P value range = 0.065–0.095) more likely, respectively, to be infected than five and six-year-old children. In contrast, sex was not associated with infection status after controlling for the other covariates.

The strong associations between infection status and tuition suggested by the single-variable models were weaker after controlling for other covariates and all confidence intervals included the possibility of no association. Depending on the geographic variable included in the model, children attending schools in the middle quartiles of tuition were as likely or slightly less likely to be infected than children attending schools in the lowest quartile of tuition. In contrast, children attending highest tuition quartile schools were up to 70% more likely to be infected than children attending lowest tuition quartile schools.

The association between nutrition program and infection status was also weakened after controlling for the other covariates and regardless of the geographic variable included in the model, confidence intervals included the null (POR = 1). Children attending schools with nutrition programs were up to 86% more likely to be infected than children from schools with no programs.

DISCUSSION

We observed very strong associations between the geographic variables (topographic zone, section, and altitude) and W. bancrofti infection status. Tuition and nutrition program effects were attenuated after controlling for the geographic variables, and confidence intervals were wide, suggesting some instability in these results. Age was modestly associated with infection status. There was no association between infection status and sex.

Ideally, we would have constructed a single model containing all three geographic variables, but the near collinearity between them prevented this. Alone, each variable provided important information, but no single variable provided a complete picture of the geographic distribution of W. bancrofti infection in Leogane Commune. The altitude findings allow for coarse identification of areas at risk for infection (areas lying below 70 meters above sea level), but some low-prevalence communities were located at lower altitudes, while the intermediate- and high-prevalence communities in sections 7 and 14 were located between 360 and 450 meters above sea level. Without the section variable, the foothill and mountain “hot spots” in sections 7, 14, and 15 might have gone unnoticed, masked by the overall low prevalence of infection observed in the mountains. Without topographic zone, the marked plains-foothills-mountains gradient of infection prevalence would not have been identified. Therefore, multiple geographic descriptors should be considered when identifying areas at risk for W. bancrofti infection.

The geographic variable results contribute in several ways to our understanding of the geographic distribution of W. bancrofti infection in Haiti, and potentially in other countries where Culex quinquefasciatus transmits LF, and may help to guide LF elimination program activities. In areas where anophelines are the predominant vectors, transmission of W. bancrofti infection at higher altitudes is assumed. Onapa and others have found W. bancrofti infection prevalences ranging from 18.3% to 30.1% in Ugandan communities located 1,000–1,100 meters above sea level.10 In areas with non-anopheline vectors, however, it is commonly accepted that W. bancrofti infection prevalence decreases sharply with increasing altitude. However, a recent study in Nepal, where Cx. quinquefasciatus is the main LF vector, found infection prevalences of 13.8–16.5% at various altitudes up to 900 meters above sea level and 9.6% between 900 meters and 1,500 meters.11 Our results indicate that in Haiti, the force of infection is greatest (and most relevant to filariasis elimination program activities) at much lower altitudes; the odds of infection drop by at least one order of magnitude above 70 meters. That said, the section results offer a caveat: although individuals living on floodplains at low altitude are much more likely to be infected than individuals living at higher altitudes, higher altitude communities are not universally exempt from infection. Filariasis elimination programs in Haiti and other areas where Cx. quinquefasciatus predominates should recognize that there may be substantial pockets of infection and potentially even transmission of infection at higher altitudes, and extend infection mapping activities into what have been assumed to be non-LF altitudes. The current results also suggest that foothills communities serve as transition areas between areas of high and low infection risk.

The characteristics, if any, shared by at-risk plains, foothills, and mountain communities remain unclear. They may share, or lack, certain landscape traits, perhaps sharing microenvironments that favor mosquito breeding, survival, and transmission of infection. Alternatively, at-risk communities may be similar with respect to the exposure opportunities of their populations; active transmission of infection might not be occurring in “unusually located” high prevalence communities themselves, but substantial proportions of their populations might be exposed elsewhere. Section 7, although in the mountains, overlooks a suburb of Port au Prince with a W. bancrofti prevalence of 7.5% (Ministère de la Santé Publique et de la Population, Haiti, unpublished data). Due to its geographic location, there is more movement between this section and a source of infection than might be expected for a mountain community. Similarly, the presence of good roads into section 14 and through section 15 results in more population movement between these sections and the plains than exists in more remote mountain settings.

Lymphatic filariasis has traditionally been considered a disease of poverty, inadequate sanitation, and underdevelopment,12–14 although the association between socioeconomic status (SES) and W. bancrofti infection has never been formally evaluated. We included tuition in the models as an indicator of the relative SES of a child’s family. School tuition may not be a precise indicator of family SES, since families value education differently; for example, some very poor families make financial sacrifices to send their children to good schools. However, to the extent that tuition is a valid marker for SES and our point estimates are stable, our results suggest that higher SES may not confer protection against infection with W. bancrofti. In this setting, children with the highest SES were, at best, as likely as children with the lowest SES to be infected. Higher SES residents of Leogane Commune may be more mobile than the less affluent, traveling frequently to urban and periurban plains areas (areas in which the force of infection and population density are highest) or settling in these more desirable areas, thus increasing their opportunity for exposure. Alternatively, all residents of a given area share the same environmental risk profile, and in this part of Haiti, higher SES does not appear to be associated with increased use of physical means of protecting oneself from vector-borne infections, such as window screens or bed nets, as it might be in other settings. Therefore, there may be no difference between the highest and lowest SES groups in terms of physical protection from W. bancrofti infection. Regardless, the tuition findings are of value to elimination programs in terms of their implications for community education and motivation and drug coverage. If SES (as defined locally) does not confer protection against W. bancrofti infection, then everyone, regardless of SES, must participate in LF elimination programs to ensure program success.

Since data collection involved only a brief encounter in the school setting with each parent-child pair, very little personal data could be collected. Consequently, there may be confounding by unmeasured covariates, particularly by individual-level characteristics such as mosquito bite prevention behaviors (e.g., bed net use, personal and household insecticide use), location of residence (urban versus rural), and home community population density. In addition, because there was no way to obtain geographic coordinates for each child’s place of residence, school location was used as a proxy for home location and values for the geographic location variables were assigned based on the location of the child’s school. However, risk of infection probably depends primarily on home location, since Cx. quinquefasciatus, the mosquito vector for W. bancrofti in Haiti, feeds at night. A review of available home community data suggested that most children lived near their schools. Nevertheless, if a child’s home and school geographies differed, there could have been non-differential misclassification of one or more geographic variables, which would bias the relevant PORs towards the null.

Our sampling methodology and the participation of all selected schools provided high resolution geographic coverage of Leogane Commune, yielding good estimates of prevalence and allowing us to discern local variations in the distribution of W. bancrofti infection. Given the clear spatial pattern suggested by the prevalence data, future work will focus on formally describing the spatial correlation between outcomes and accounting for this correlation when modeling infection status. In addition, we aim to determine whether at-risk locations share common landscape characteristics by assessing the relationship between landscape traits, region, and W. bancrofti infection using remote sensing data and these school data.

Table 1

Study population school and student characteristics (Leogane Commune, Haiti, October 1999–June 2000 and March–May 2001)

Number of schools (%)Number of children (%)
* The ages of 19 children were unknown.
† The rate of exchange in 1999–2000 was 17 gourdes to one US dollar.
‡ One school (68 children) was missing tuition information.
§ Two schools (37 children) were missing nutrition program information.
Age (years)*
    5NA393 (11.9)
    6NA419 (12.7)
    7NA502 (15.2)
    8NA530 (16.1)
    9NA462 (14.0)
    10NA499 (15.1)
    11NA494 (15.0)
Sex
    MaleNA1,679 (50.6)
    FemaleNA1,639 (49.4)
Tuition (gourdes/year)†‡
    <10014 (25.0)845 (26.0)
    100–19917 (30.4)787 (24.2)
    200–29913 (23.2)963 (29.6)
    300–3993 (5.4)183 (5.6)
    400–4995 (8.9)314 (9.7)
    ≥5004 (7.1)158 (4.9)
Nutrition program§
    Program in place20 (36.4)1,525 (46.5)
Topographic zone
    Mountains22 (38.6)1,159 (34.9)
    Foothills13 (22.8)812 (24.5)
    Plains22 (38.6)1,347 (40.6)
Section
    113 (22.8)747 (22.5)
    25 (8.8)414 (12.5)
    38 (14.0)430 (13.0)
    67 (12.3)311 (9.4)
    73 (5.3)97 (2.9)
    81 (1.8)68 (2.0)
    91 (1.8)75 (2.3)
    104 (7.0)227 (6.8)
    112 (3.5)71 (2.1)
    123 (5.3)280 (8.4)
    132 (3.5)82 (2.5)
    144 (7.0)271 (8.2)
    154 (7.0)245 (7.4)
Altitude (meters above sea level)
    ≤2518 (31.6)1,162 (35.0)
    26–7513 (22.8)698 (21.0)
    76–1504 (7.0)299 (9.0)
    151–4009 (15.8)542 (16.5)
    >40013 (22.8)617 (18.6)
Table 2

Distribution of Wuchereria bancrofti infection with respect to individual- and school-level characteristics and single-variable logistic generalized linear model results (Leogane Commune, Haiti, October 1999–June 2000 and March–May 2001)

VariableTotal number of children testedNumber infected (%)POR* (95% CI)
* Unadjusted prevalence odds ratio (POR) (95% confidence interval [CI]) obtained from single-variable logistic generalized linear mixed model. Interpreted given the school each child attended.
† 1st quartile: tuition ≤90 gourdes/yr; 2nd quartile: 90 gourdes/yr < tuition ≤182.5 gourdes/ yr; 3rd quartile: 182.5 gourdes/yr < tuition ≤ 275 gourdes/yr; 4th quartile: tuition >275 gourdes/yr.
‡ Refer to Leogane Commune map (Figure 1) for section locations.
§ 1st quartile: altitude ≤15 m; 2nd quartile: 15 m < altitude ≤70 m; 3rd quartile: 70 m < altitude ≤ 400 m; 4th quartile: altitude > 400 m.
Age (years)
    5 and 6812117 (14.4)1.00 (Referent)
    7 and 81,032179 (17.3)1.29 (1.01, 1.66)
    9 to 111,455236 (16.2)1.26 (0.99, 1.61)
Sex
    Female1,679244 (14.5)1.00 (Referent)
    Male1,639289 (17.6)1.15 (0.96, 1.38)
Tuition†
    1st Quartile845175 (20.7)1.00 (Referent)
    2nd Quartile66359 (8.9)0.71 (0.17, 3.05)
    3rd Quartile1,023124 (12.1)0.55 (0.13, 2.29)
    4th Quartile719175 (24.3)2.84 (0.68, 11.81)
Nutrition program
    No program1,756265 (15.1)1.00 (Referent)
    Program in place1,525259 (17.0)2.26 (0.76, 6.72)
Topographic zone
    Mountains1,15928 (2.4)1.00 (Referent)
    Foothills81284 (10.3)4.26 (1.56, 11.61)
    Plains1,347421 (31.1)28.95 (12.26, 68.37)
Section‡
    1747246 (32.9)53.34 (14.63, 194.49)
    2414154 (37.2)69.69 (17.38, 279.52)
    343091 (21.2)26.71 (7.04, 101.36)
    6, 8, and 94543 (0.7)0.70 (0.13, 3.84)
    79712 (12.4)16.17 (3.45, 75.71)
    10 and 133093 (1.0)1.00 (Referent)
    11 and 123513 (0.9)0.89 (0.15, 5.10)
    1427111 (4.1)3.21 (0.68, 15.19)
    1524510 (4.1)4.69 (1.00, 22.03)
Altitude quartiles§
    1st1,104323 (29.3)1.00 (Referent)
    2nd756178 (23.5)0.83 (0.35, 1.96)
    3rd84119 (2.3)0.04 (0.02, 0.13)
    4th61713 (2.1)0.04 (0.01, 0.13)
Alternate altitude ranges (meters)
    Altitude <251,104323 (29.3)1.00 (Referent)
    25 ≤ altitude ≤ 50469145 (30.9)1.39 (0.51, 3.77)
    50 < altitude ≤ 15058637 (6.3)0.16 (0.05, 0.45)
    150 < altitude ≤ 40054215 (2.8)0.05 (0.01, 0.16)
    Altitude > 40061713 (2.1)0.04 (0.01, 0.14)
Table 3

Prevalence odds ratios for geographic variables from multivariate logistic generalized linear mixed models*

VariablePOR† (95% CI)
* Adjusting for age, sex, tuition, and nutrition program.
† Prevalence odds ratio (POR) (95% confidence interval [CI]). All prevalence odds ratios are interpreted given the school each child attended.
‡ 1st quartile; altitude ≤ 15 m; 2nd quartile; 15 m < altitude ≤70 m; 3rd quartile; 70 m < altitude ≤ 400 m; 4th quartile; altitude > 400 m.
Topographic zone
    Mountains1.00 (Referent)
    Foothills3.95 (1.28, 12.23)
    Plains23.56 (8.99, 61.79)
Section
    162.06 (15.59, 247.12)
    246.09 (10.81, 196.52)
    326.82 (6.53, 110.09)
    6, 8, and 90.89 (0.15, 5.26)
    730.49 (5.77, 161.18)
    10 and 131.00 (Referent)
    11 and 120.97 (0.15, 6.35)
    143.11 (0.62, 15.63)
    153.10 (0.55, 17.50)
Altitude quartiles‡
    1st1.00 (Referent)
    2nd0.80 (0.31, 2.05)
    3rd0.04 (0.01, 0.12)
    4th0.07 (0.02, 0.23)
Alternate altitude ranges (meters)
    Altitude <251.00 (Referent)
    25 ≤ altitude ≤ 501.27 (0.42, 3.84)
    50 < altitude ≤ 1500.17 (0.05, 0.53)
    150 < altitude ≤ 4000.04 (0.01, 0.14)
    Altitude > 4000.07 (0.02, 0.23)
Figure 1.
Figure 1.

School sampling in Leogane Commune, Haiti, October 1999–June 2000 and March–May 2001. CDC = Centers for Disease Control and Prevention.

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 70, 3; 10.4269/ajtmh.2004.70.266

Figure 2.
Figure 2.

Wuchereria bancrofti infection prevalence in Leogane Commune, Haiti by student home community (October 1999–June 2000 and March–May 2001).

Citation: The American Journal of Tropical Medicine and Hygiene Am J Trop Med Hyg 70, 3; 10.4269/ajtmh.2004.70.266

Authors’ addresses: Heather A. Boyd, Department of Epidemiology, Rollins School of Public Health, Emory University, Room 468, 1518 Clifton Road, Atlanta, GA 30322 and Division of Parasitic Diseases, Centers for Disease Control and Prevention, Chamblee Facility, Mailstop F-22, Building 24, Room 1006, 4770 Buford Highway, Atlanta, GA 30341-3717, Telephone: 770-488-7595, Fax: 770-488-4465, E-mail: heb6@cdc.gov. Lance A. Waller, Department of Biostatistics, Rollins School of Public Health, Emory University, Room 326, 1518 Clifton Road, Atlanta, GA 30322, Telephone: 404-727-1057, Fax: 404-727-1370, E-mail: lwaller@sph.emory.edu. W. Dana Flanders, Department of Epidemiology, Rollins School of Public Health, Emory University, Room 464, 1518 Clifton Road, Atlanta, GA 30322, Telephone: 404-727-8716, Fax: 404-727-8737, E-mail: wflande@sph.emory.edu. Michael J. Beach, Centers for Disease Control and Prevention, Chamblee Facility, Building 102, Room 1411, Mailstop F-22, 4770 Buford Highway, Atlanta, GA 30341-3717, Telephone: 770-488-7763, Fax: 770-488-7761, E-mail: mjb3@cdc.gov. J. Sony Sivilus and Rodrigue Lovince, Hôpital Sainte Croix, Leogâne, Haiti, Telephone: 509-235-1044; E-mails: jeanssivilus@yahoo.fr and Rodrigue.Lovince.1@nd.edu. Patrick J. Lammie, Centers for Disease Control and Prevention, Chamblee Facility, Building 23, Room 1021, Mailstop F-22, 4770 Buford Highway, Atlanta, GA 30341-3717, Telephone: 770-488-4054, Fax: 770- 488-4108, E-mail: pjl1@cdc.gov. David G. Addiss, Centers for Disease Control and Prevention, Chamblee Facility, Building 24, Room 1007, Mailstop F-22, 4770 Buford Highway, Atlanta, GA 30341-3717, Telephone: 770-488-7770, Fax: 770-488-4465, E-mail: dga1@cdc.gov.

Acknowledgments: We thank Joyanna Wendt and Jeanne Radday (Leogane Commune LF elimination program coordinators at the time this work was undertaken), Dr. Jack Lafontant (Director, Hôpital Sainte Croix), and Dr. Thomas Streit for their support during the school survey data collection. Dr. David Molyneux’s support and enthusiasm for the project were also much appreciated. We are very grateful to Allen Hightower for his invaluable advice and technical assistance, without which the GPS mapping work could not have been accomplished. We also thank Dr. Brian Stephens for the two months of hard work required to create the GPS base map of the Commune, and Renn Doyle for digitizing the only existing map of the Commune. We are indebted to the following elimination program staff members for their dedication to the program and for their willingness to travel the length and breadth of the Commune during data collection activities: Conus Brevaus, Jean Marc Brissau, Dardith Desire, Erubens Elaus, Shiler Emile, Fred Francis, Jean Steven Hilaire, Jean Willy Hilaire, Jesnel Jorseme, Michelet Leriche, Jean Clausel Louis, Rose Guertha Louis-Charles, Wesly Pierre, Osnel Sagaille, and Kethlie Toussaint.

Financial support: This work was supported by the Department for International Development, United Kingdom, through the Liverpool School of Tropical Medicine Lymphatic Filariasis Support Centre and by the CDC Emerging Infections Program.

REFERENCES

  • 1

    Michael E, Bundy DAP, Grenfell BT, 1996. Re-assessing the global prevalence and distribution of lymphatic filariasis. Parasitology 112 :409–428.

    • Search Google Scholar
    • Export Citation
  • 2

    World Health Organization, 1997. World Health Assembly Resolution 50.29. Geneva: World Health Organization.

  • 3

    Ramzy RM, Hafez ON, Gad AM, Faris R, Harb M, Buck AA, Weil GJ, 1994. Efficient assessment of filariasis endemicity by screening for filarial antigenaemia in a sentinel population. Trans R Soc Trop Med Hyg 88 :41–44.

    • Search Google Scholar
    • Export Citation
  • 4

    World Health Organization, 1998. Research of Rapid Geographic Assessment of Bancroftian Filariasis. Geneva: World Health Organization. TDR/TDF/COMDT/98.2).

  • 5

    World Health Organization, 1999. Report of a WHO Informal Consulation on Epidemiologic Approaches to Lymphatic Filariasis Elimination: Initial Assessment, Monitoring and Certification. Geneva: World Health Organization. WHO/FIL/99.195.

  • 6

    World Health Organization, 1999. Guidelines for Certifying Lymphatic Filariasis Elimination (Including Discussion of Critical Issues and Rationale). Geneva; World Health Organization. WHO/FIL/99.197.

  • 7

    Breslow NE, Clayton DG, 1993. Approximate inference in generalized linear mixed models. J Am Stat Assoc 88 :9–25.

  • 8

    Agresti A, Booth JG, Hobert JP, Caffo B, 2000. Random-effects modeling of categorical response data. Sociol Methodol 30 :27–80.

  • 9

    Littell RC, Milliken GA, Stroup WW, Wolfinger RD, 1996. SAS® System for Mixed Models. Cary, NC: SAS Institute, Inc., 423–460, 491–495.

  • 10

    Onapa AW, Simonesen PE, Pedersen EM, Okello DO, 2001. Lymphatic filariasis in Uganda: baseline investigations in Lira, Soroti and Katakwi districts. Trans R Soc Trop Med Hyg 95 :161–167.

    • Search Google Scholar
    • Export Citation
  • 11

    Sherchand JB, Obsomer V, Thakur GD, Hommel M, 2003. Mapping of lymphatic filariasis in Nepal. Filaria J 2 :7. (http://www.filariajournal.com/content/2/1/7).

    • Search Google Scholar
    • Export Citation
  • 12

    Molyneux DH, Zagaria N, 2002. Lymphatic filariasis elimination: progress in global programme development. Ann Trop Med Parasitol 96 :S15–S40.

    • Search Google Scholar
    • Export Citation
  • 13

    Evans DB, Gelband H, Vlassoff C, 1993. Social and economic factors and the control of lymphatic filariasis: a review. Acta Trop 53 :1–26.

    • Search Google Scholar
    • Export Citation
  • 14

    Hunter JM, 1992. Elephantiasis: a disease of development in northeast Ghana. Soc Sci Med 35 :627–649.

Author Notes

Reprint requests: David G. Addiss, Centers for Disease Control and Prevention, Chamblee Facility, Mailstop F-22, 4770 Buford Highway, Atlanta, GA 30341-3717.
Save