Giardia duodenalis infection and anthropometric status in preschoolers in Salvador, Bahia State, Brazil Infecção por Giardia duodenalis e estado antropométrico em pré-escolares de Salvador, Bahia, Brasil

The aim of this study was to estimate the association between Giardia duodenalis infection and anthropometric deficits, as measured by weight-for-age and height-for-age. This crosssectional study included 629 children from 12 to 48 months of age, selected from 30 geographic areas in the city of Salvador, Bahia State, Brazil. Poisson regression and linear regression were used for the multivariate statistical analyses. G. duodenalis was diagnosed in 13.5% of the children. The children’s breastfeeding duration and living conditions (garbage collection and paved streets or sidewalks) modified the effect of G. duodenalis infection on anthropometric status. Among infected children, there were statistically significant associations between weight deficit and shorter breastfeeding (PR = 2.22; 95%CI: 1.56-3.14) and inadequate paving of streets and sidewalks (PR = 2.00; 95%CI: 1.37-2.92), while height deficit was associated with deficient public garbage collection (PR = 2.21; 95%CI: 1.312.51). In the linear regression, the association with the anthropometric indicators remained positive and statistically significant. The child’s unhealthy living environment aggravated the negative effect of G. duodenalis infection on anthropometric status, and breastfeeding was a protective factor in the outcome. Giardia; Anthropometry; Preschool Child Introduction Malnutrition and intestinal parasites, especially Giardia duodenalis infection, are found simultaneously in underdeveloped regions, and the bidirectional relationship between these two events is well known 1. Most studies investigating the influence of intestinal parasites on malnutrition are limited to regions where the two are highly prevalent 2,3,4,5. However, even where moderate and severe malnutrition has been declining, as in the area where this study was conducted 6, G. duodenalis infection can be expected to exert a limiting impact on childhood growth 7. The negative impact of G. duodenalis infection on anthropometric status can be explained by jeopardized food intake due to anorexia and poor absorption of macro and micronutrients. The infection can thus have a worse impact on preschool-age children, in whom nutritional needs are higher due to the intense linear growth and weight gain in this phase of life 8,9. Although G. duodenalis infection does not pose a direct risk of death in childhood, its association with malnutrition and diarrhea makes it an important health problem, and it can contribute indirectly to increased mortality rates in childhood 10. In the preschool population in developing countries, the incidence of G. duodenalis infection is higher than in industrialized countries 11 and is heavily influenced by the children’s surARTIGO ARTICLE Matos SMA et al. 1528 Cad. Saúde Pública, Rio de Janeiro, 24(7):1527-1535, jul, 2008 rounding socioeconomic and sanitary conditions 4,12. Thus, low-income populations living with precarious sanitation are the most affected. In this context, inappropriate disposal of excreta, insufficient drainage of surface water, and scarce or inadequate water supply favor the dispersal of pathogens in the environment, especially when added to nonexistent or inadequate personal and household hygiene. Precarious basic sanitation further aggravates the risk of infectious and parasitic diseases, thereby negatively influencing childhood nutritional status. Thus, malnutrition and intestinal parasitic infections coexist in poor regions with adverse economic conditions. Considering the limited exploration of this association in studies focusing on urban areas permeated by overt social, economic, and environmental disparities, the current article proposes to verify the association between G. duodenalis infection and anthropometric status in preschoolers in Salvador, Bahia State, Brasil. Material and methods The study adopted a cross-sectional design to investigate the relationship between G. duodenalis infection and childhood anthropometric status, within the context of a broader study aimed at evaluating the Epidemiological Impact of the Environmental Sanitation Program in the All Saints’ Bay – Blue Bahia, in Salvador 13,14. As part of this evaluation project, in the year 1997 a group of 30 geographic (sentinel) areas were selected in the city, located in ten different sewage basins. These areas represented the diversity of sanitary and socioeconomic conditions in the city, and they were chosen from 1,765 census tracts in Salvador based on the 1991 population census and totaling approximately 20 thousand households. The census tracts were classified in three different levels according to sewage disposal standards and family income. This procedure has been described elsewhere 15. In each of these areas, households were selected with children less than three years of age. The current study included 629 children 48 months of age or younger that presented complete information in the variables used here. The study excluded children with discrepant values for the anthropometric indicators, as recommended by the World Health Organization (WHO) 16, based on the margins set for mean Z-score. The sample thus excluded cases with Z-scores of < -5.0 and > +3.0 for height-for-age (H/A), < -5.0 and > +5.0 for weight-for-age (W/A), and < -4.0 and > +5.0 for weight-for-height (W/H). Mild malnutrition was defined as mean Zscores less than -1 SD Z-score. Due to the reduced prevalence of moderate/severe anthropometric deficits (< -2 SD Z-score), it was decided to include mild malnutrition as a nutritional risk condition. The inclusion of mild malnutrition in the analyses was further justified by the intense discussion of the relevance of this form of malnutrition for childhood mortality 17,18,19,20. Weight and height/length measurements were used to evaluate the children’s anthropometric status. The evaluation was performed from June to September 1998 in strategic sites previously scheduled for each residential area. The children were weighed on portable microelectronic scales with a capacity of 150kg and accurate to 100g (model E-150/3P, Filizola Balanças Industriais, São Paulo, Brazil), ceded by the defunct National Institute of Food and Nutrition/ Ministry of Health (INAN). Children younger and older than 24 months, respectively, were measured with wooden infantometers and stadiometers (Leicester Height Measure, Seca, Hamburg, Germany). The instruments were calibrated periodically and were replaced with inelastic tape measures when necessary. All the measurements were taken in tandem, by properly trained nutritionists and nutrition students, allowing a variation of 100g and 0.1cm, respectively, for weight and length/height, and the mean of the two was used as the final measurement 21. The recommended standards 16 and technical criteria 21 were observed in all stages of the anthropometric evaluation. Age was recorded from the birth certificate or child’s health card, and the difference was calculated between the measurement date and the child’s date of birth. The study used ANTHRO – Software for Calculating Pediatric Anthropometry (Centers for Disease Control and Prevention, Atlanta, USA), which uses the National Center for Health Statistics standard 22 as the reference for deriving Z-scores for H/A, W/A, and W/H. The parasitological survey was conducted in parallel with the anthropometric evaluation, and a single stool sample was taken for each child. The child’s parent or guardian received at home a specific stool-test recipient with a lid, labeled and numbered, and was oriented on how to collect the stool sample in this recipient or directly from the diaper into the recipient. Stool samples were collected the day after the recipients were distributed, and two attempts were made when the specimen was not available on the scheduled day. When this criterion was not met, the child was considered lost to the sample. After the stool sample was taken, the specimens were placed in cold storGiardia duodenalis INFECTION AND ANTHROPOMETRIC STATUS IN PRESCHOOLERS 1529 Cad. Saúde Pública, Rio de Janeiro, 24(7):1527-1535, jul, 2008 age, transported to the laboratory, and examined the same day. Helminth diagnosis and fecal egg count used the Kato-Katz method 23 and G. duodenalis cysts were identified using the spontaneous sedimentation method 24. According to the number of eggs excreted in the feces, the intensity of Ascaris lumbricoides infection was classified as mild (< 5,000 eggs/g), moderate (5,000-50,000 eggs/g), or heavy (≥ 50,000 eggs/g) and Trichuris trichiura infection as mild (≤ 1,000 eggs/g), moderate (1,000-10,000 eggs/g), or heavy (≥ 10,000 eggs/g) 25. The mean number of eggs per gram of feces was also calculated. Demographic, socioeconomic, and environmental data were obtained with a semi-structured, pre-coded questionnaire. Information on behavior standards for household and personal hygiene was obtained from the child’s parent or guardian, and the behavior in which the child was involved was observed and recorded after each visit. This pattern was based on 34 behaviors related to the child or parent/guardian, which were grouped into 16 negative and 18 positive behaviors and situations. The standardized sum of these behaviors allowed classifying the sample into groups with “positive”, “intermediate”, and “negative behavior”. More detail on this classification can be found in Strina et al. 14 and Barreto et al. 26. Prevalence of G. duodenalis infection was adopted as the measure of occurrence and compared between the exposure groups using prevalence ratio (PR) as the measure of association. Since the study outcome is not a rare event, overestimation of the odds ratio may be observed as compared to the prevalence ratio. Therefore, rather than using logistic regression, multivariate Poisson regression models were defined with robust variance 27,28, considering G. duodenalis infection as the principal exposure and the H/A and W/A indicators as dependent variables. Effect modification was interpreted as change in the magnitude of the effect in the presence of a third variable 29,30. Wald test was used to evaluate interaction between an exposure variable and potential effect-modifying co-variables. “Interaction variables” were defined as those in which the χ2 reached a p-value < 0.10. The statistical criterion adopted for recognizing a confounder was a shift of at least ten percentage points in the measurement of association caused by the variable as compared to that estimated in its absence 29,30. The backward progressive elimination procedure was used to perform these statistics. Multicolinearity between the independent variables was evaluated by means of correlation coefficients, and the absence of this condition was identified when the correlation coefficient was less than 0.35 31. Given the characteristics of the sample calculation for the original study (cluster-sampled), the existence of interdependence was assumed between the measurements generated for children from the same sentinel area. Thus, it was necessary to weight the statistical analysis to adjust the standard errors to the intra-group correlation. The “jackknife” estimation method was used for this purpose, comparing the sentinel areas by executing the command for each area omitted from an observation in turn (Stata Statistical Software, Stata Corp., College Station, USA). In addition to the use of the multivariate Poisson regression model, the choice was made for an alternative analytical strategy meeting the model’s premises, namely the use of a multiple linear regression model with a robust estimate of variance. The hierarchical approach was used for entering the variables in the blocks 32, initially adopting a statistical significance of 0.20 to select possible variables to be included in the multivariate model and a significance of 0.10 to select the variables that would remain in the model. Variables related to the child’s factors (level 2) were added to this model, and the new analysis was conducted, maintaining in the model the variables with significance of 0.10. The final model for predicting anthropometric deficit consisted of the variables selected in each of the stages. Epi Info, version 6.4 (Centers for Disease Control and Prevention, Atlanta, USA) was the statistical package adopted for data entry, and the analyses were performed using Stata, version 9.0. The study was approved by the Institutional Review Board of the Institute of Collective Health, Federal University in Bahia. Results of the stool tests (both positive and negative) were returned to the children’s parents. In case of positive test results, the children received the appropriate treatment. When necessary, nutritional orientation was provided to the child’s parent or guard-