Although cardiovascular disease mortality has declined substantially over the past three decades in most developed countries, marked disparities persist for both mortality and risk factor prevalence by socioeconomic status (Kawachi et al. 1991, Rose and Marmot 1981, Shea et al. 1991). Low socioeconomic status groups consistently show a higher prevalence of smoking, hypercholesterolemia (Luepke et al. 1993), hypertension (Luepke et al. 1993) and leisure time physical inactivity (Rose and Marmot, 1981, Shea et al. 1991) than more affluent groups. In addition, several dietary surveys indicate that poverty is associated with low intakes of vitamin C (Simon et al. 1993), folate (Ballew and Sugerman 1995), and high intakes of total and saturated fat (Ballew and Sugerman 1995, Devaney et al. 1995, Laitinen et al. 1995).

Surveys of school children report higher intake of fat, lower intake of complex carbohydrates and lower intake of some micronutrients among those from poor families (Devaney et al. 1995, Laitinen et al. 1995), but this is not a consistent finding (Johnson et al. 1994). However, obesity occurs more frequently among children of low socioeconomic status (Jones et al. 1985, Rolland-Cachera and Bellisle 1986, Yip et al. 1993) than in the general population. In addition, over the past two decades, a secular trend toward increased obesity in children (Gortmaker et al. 1987, Limbert et al. 1994) is more marked among children of low socioeconomic status than among those more advantaged (Jones et al. 1985, Yip et al. 1993) and varies by ethnicity or race (Troiano et al. 1995). Of concern is that obesity acquired in childhood "tracks" or persists into adulthood (Berenson et al. 1994, Clarke and Lauer 1993, Dietz 1987), and unhealthy lifestyle habits, including dietary habits learned early in life, are thought to be more difficult to change when habits are ingrained (Berenson et al. 1994).

Although it is clear that low income children have a higher risk of obesity (Yip et al. 1993) and may consume a less "prudent diet" (Devaney et al. 1995, Johnson et al. 1994), no data encompassing obesity, diet and activity are available on the nutritional status of children from poor multiethnic families. As part of a heart health promotion program, we studied the dietary intake and anthropometric indices of 9- to 12-y-old children from poor, multiethnic, inner-city neighborhoods in Montreal, Canada. The objectives were to assess the prevalence of obesity and/or undernutrition, to explore issues of energy balance and describe the quality and quantity of dietary intake as related to indices of poverty, ethnicity and other family and personal characteristics.

SUBJECTS AND METHODS

In May/June 1993, 2285 of 2840 eligible students (80.5%) aged 9-13 y in all 130 Grade 4, 5 and 6 classes in the 24 participating schools completed a base-line survey. Data on sociodemographic characteristics, level of physical activity and other lifestyle habits were collected in self-administered questionnaires that were started at school and completed by students at home. Data on income sufficiency, and mother's and father's education were collected in self-administered questionnaires completed by the subjects' parents. Anthropometric measures of height and weight were obtained by trained interviewers according to a standardized protocol (Evers and Hooper 1995) on the day of the in-class questionnaire.

This base-line sample of 2285 students comprised the sampling frame for this study on the nutritional status of this population. Five students in each of the 130 classrooms were randomly selected from the list of students who had agreed to participate in the study, and each student was personally interviewed to obtain a 24-h dietary recall. A total of 563 dietary interviews were conducted.

Frequency of physical activity was determined in an adaptation of the self-reported Weekly Activity Checklist, a 7-d recall of physical activity described by Sallis and colleagues (1993). Briefly, students checked which of 28 different physical activities they had completed for each day during the week preceding the interview. The list of activities had been determined during extensive pretesting of the instrument and included those activities most frequently engaged in by this age group during the spring months. A frequency score was computed for each student by summing the total number of activities checked for each day. The median number of activities was 12, with a 25th percentile of 7 and 75th percentile of 21.

Body mass index (BMI)⁵ was computed by weight/height² (kg/m²). Students were categorized according to age and gender-specific percentile of BMI using NHANES II standards (Frisancho 1990). Underweight was defined as less than or equal to the age and gender-specific 15th percentile. Normal weight was defined as between the 15th and 85th age and gender-specific percentile. Because students near the 85th percentile might be muscular or have a large bone structure rather than be overweight, we categorized students between the 85th and 90th age and gender-specific percentile as somewhat overweight. Moderately overweight was defined as between the 90th and 95th age and gender-specific percentile; very overweight included students over the age and gender-specific 95th percentile. Must et al. (1991) have indicated that the 85th and 95th percentiles for BMI do not exhibit racial differences in preadolescent children in the United States. Recently, BMI values above the 85th percentile have been shown to predict adiposity using dual-energy X-ray absorptiometry (Lazarus et al. 1996).

Dietary intake was measured by a single dietitian-administered 24-h recall per child using graduated food models of food portion sizes (Santé Quebec, Montreal, Canada). These models, used in the recent Quebec Nutritional Survey (Santé Quebec 1995) included a variety of shapes (e.g., spheres or ellipses) of different sizes with which to compare food portions, a technique suggested for children by Frank et al. (1977). To help children remember what they had eaten, each student completed a four-page dietary record/checklist, developed by members of the study team, for the 24-h period preceding the interview. Students were encouraged to obtain help from their parents to complete the record. Dietitians conducted the 24-h dietary recalls unaware of the contents of the students' record/checklist. Because of the difficulty collecting data on weekends, only data for Monday to Thursday were collected. The dietary interview was done on the school day following the administration of the demographic and lifestyle questionaire and the anthropometric measurements.

To study possible underreporting of dietary intake by young children, we calculated the ratio of reported energy intake (EI) to basal energy requirements in a resting and fasting state (Black et al. 1991). Basal metabolic rate (BMR _est) was calculated using both actual and ideal body weight (Black et al. 1991, FAO/WHO 1985) because overweight children may have the same fat-free mass and BMR as normal weight children of comparable height (Delany et al. 1995). In adults, a mean ratio of EI:BMR_est < 1.5 for a group indicates underreporting (Black et al. 1991). Because estimated energy expenditure in children is higher than in adults, a ratio of < 1.8 was considered to indicate underreporting in children, based on observed ratios of total energy expenditure to BMR_est calculated on the basis of age, gender and body weight (Livingstone et al. 1992).

Quality of diet was assessed by comparing the percentage of fat and grams of fiber in the diet and mean daily intake of micronutrients at risk of deficiency in children, including vitamins A and C, calcium, iron, zinc and folate to the Canadian and American recommendations (Health and Welfare Canada 1990, NRC 1989) (Johnson et al. 1994). The associations between mean daily intake of each nutrient of interest and each of age, gender, income sufficiency, number of persons in household, level of physical activity and mother's country of birth were studied with the use of one-way ANOVA. Differences in mean levels of nutrients were identified in pairwise comparisons by using the Scheffé procedure (Armitage and Berry 1994). A multivariate logistic regression analysis was done to identify independent correlates of obesity (>85th percentile, Frisancho 1990). All analyses were conducted with the use of the SAS statistical package (SAS Version 6.0, SAS Institute, Cary, NC).

RESULTS

Table 1. Sociodemographic and anthropometric characteristics of children in multiethnic, low income, inner-city neighborhoods, Montreal, Canada [View Table]

The proportion of overweight children was high, i.e., 41.8% of boys and 36.9% of girls were above the 85th BMI percentile (Table 2). Compared with the NHANES II reference population, there were 2.6 times more overweight children in our low income multiethnic sample than expected. Half of the overweight children were very heavy (>95th percentile). Only 9.7% of the sample were underweight, lower than the expected 15%. We observed that 51% of the study subjects were in the normal range (15th-85th percentiles) where 70% was expected.

Table 2. Percentage of children in multiethnic, low income, inner-city neighborhoods by body mass index, Montreal, Canada [View Table]

The mean EI:BMR_est in this study was 1.79 ± 0.71, indicating no serious underreporting overall. There was no difference in EI:BMR_est by age or gender. However, children in the highest BMI quartile showed evidence of significant underreporting; their mean value was 1.31 ± 0.45 (Table 3). The pattern of decreasing EI:BMR_est with increasing weight was evident in both boys and girls. Reported activity levels did not differ between weight groups, suggesting that the lower energy intake in overweight children was not attributable to lower activity levels. When assessed using ideal rather than actual body weight to compensate for the possibility that the use of actual body weight may overestimate BMR, the EI:BMR_est for the highest BMI quartile was 1.56 ± 0.55, still significantly lower than the other BMI quartiles.

Table 3. Ratio of energy intake (EI) to basal metabolic rate (BMR) by body mass index (BMI) among children in multiethnic, low income, inner-city neighborhoods, Montreal, Canada1 [View Table]

Quality of diet was assessed by percentage of fat in the diet and by mean daily intake of important vitamins and minerals. In Table 4, nutrient comparisons by age and gender, income sufficiency, household size, mother's country of origin and physical activity are shown. Total intakes of energy, iron and zinc were significantly lower in girls compared with boys (P < 0.01). Although energy intake increased with age among the boys, no such increase was observed for the girls. Overall, fat consumption was not high: the mean dietary fat intake as a percentage of energy was 30%. With the exception of calcium in 10- to 12-y-old girls, the mean intake of all nutrients studied met both the Canadian Recommended Nutrient Intakes (RNI) and the U.S. Recommended Daily Allowances (RDA) (Health and Welfare Canada 1990, NRC 1989). There was no association between income sufficiency and total energy intake. However, lower income sufficiency groups had significantly lower levels of vitamins A and C, iron and folate. None of these mean values, however, fell below the RNI or RDA. The number of persons per household was inversely associated with percentage of energy as fat as well as calcium. Furthermore, in a separate comparison of mean values, children in single-parent families had significantly higher intakes of percentage of energy as fat (32% vs. 29%, P < 0.001) and higher calcium intakes (1113 vs. 979 mg, P = 0.03).

Table 4. Mean daily intake of selected nutrients by age and gender among children in multiethnic, low income, inner-city neighborhoods, Montreal, Canada1 [View Table]

BMI was not related to income sufficiency or to household size. In a multivariate logistic regression analysis of the correlates of obesity, age, sex, number of children in the household, maternal education, mothers' country of origin and activity level were not related to obesity. The relationship of ethnic origin to obesity has been studied more extensively in the larger sample in which the number of subjects from each region was more suitable for this analysis (J. O'Loughlin, Montreal General Hospital; G. Paradis, L. Renaud, G. Meshefedjian, K. Gray-Donald, unpublished data). It was not appropriate to study the relationship of diet and BMI given the underreporting among the heavy children.

Children whose mothers were born in Canada consumed the highest mean intake of percentage energy as fat; the differences were significant compared with those whose mothers were born in Asia, Central and South America. Vitamin A intake was significantly lower among subjects whose mothers were born in Central or South America, compared with those whose mothers were born in Canada. Neither total energy intake nor BMI varied by the regions of mothers' country of birth.

Because of the high prevalence of obesity, we studied the association between frequency of physical activity (grouped by quartiles) and dietary intake (Table 4). Higher frequency of physical activity was related to significantly higher mean intakes of energy, calcium, iron, zinc and fiber. Despite the higher absolute amount of these nutrients, there was no difference in the macronutrient composition of the diet when a higher activity level was reported. BMI was not associated with frequency of physical activity so that the association between activity and nutrient intake was unlikely to have been confounded by this variable.

DISCUSSION

It is always important to assess the reporting accuracy of children. Our overall energy intake results compare favorably with recent large dietary studies of children in Canada and the U.S. (Devaney et al. 1995, Nicklas et al. 1993, Seoane and Roberge 1983), but our results are the first to suggest underreporting among overweight children. The systematic underreporting of overweight individuals has been observed previously in adults (Lichtman et al. 1992) and adolescents (Bandini et al. 1994, Livingston et al. 1992), but it was not clear at what age this problem first occurs. This underreporting among children as young as 9 y may indicate that these children already perceive what are socially desirable responses to dietary interviews. Such a systematic effect makes it difficult to evaluate whether diets are different between normal and overweight children. In addition, overweight children did not report doing less physical activity than normal weight children. Both conditions make it difficult to clearly identify diet or physical inactivity as the primary source of their energy imbalance.

A clear pattern of increasing nutrient intake with increasing activity emerged in our study. A greater frequency of physical activity was associated with an increased dietary intake of total energy, calcium, iron, zinc and fiber. Although studies have examined energy intake and physical activity as correlates of obesity (Gazzaniga and Burns 1993, Obarzanek et al. 1994), no studies had previously established the relationship between higher nutrient intake and higher physical activity levels in this age group. Higher energy intakes (Adamson et al. 1992, Gliksman et al. 1993) and/or fat intakes (Devaney et al. 1995, Gliksman et al. 1993) have been reported in children of lower socioeconomic status. Our data do not confirm this finding. Despite the high prevalence of obesity, the children consumed a balanced diet. Dietary fat intake in our study was lower than that reported in earlier dietary studies (Johnson et al. 1994, Nicklas et al. 1993) but very similar to the most recent study in a low income population (Evers and Hooper 1995). Additionally, the diets of immigrant Canadian children had a lower intake of fat as percentage of energy than children whose mothers were born in Canada; significantly lower fat intake was observed for those born in Asia or Central and South America.

This is the first study to measure correlates of diet and obesity with income, ethnicity and household size. The only nutrient to be deficient in our study was calcium among girls, a finding consistent with other studies of Canadian (Mailhot et al. 1994) and American children (Devaney et al. 1995). In this low income multiethnic school sample, several other important nutrients followed an income gradient (vitamins A, C, iron and folate). This observation that children in the poorer families had lower consumption of certain nutrients found widely in fruits and vegetables concurs with a report by Adamson et al. (1992). Children living in large households consumed a lower percentage of energy as fat and lower intakes of calcium. The only other study to compare intakes by household size, Johnson et al. (1994), did not find lower levels of nutrients by household size. Although lower risk of cardiovascular disease has been attributed to lower fat intakes in some ethnic groups migrating to the U.S. (Egusa et al. 1993, Gliksman et al. 1993), no dietary data have existed specifically for school age children.

This study provides nutritional data on inner city, low income children. Studies of low income groups have often noted low response rates, possibly compromising external validity (Johnson et al. 1994). Despite our high response rate (80.5%), one cannot rule out the possibility that those families that refused to have their children included in the survey may be those with underweight children or a poorer quality diet. In addition, Miller and Korenman (1994) indicate that despite finding no elevated risk of undernutrition among U.S. children from poor families in a cross-sectional analysis, they found both stunting and wasting in children living in persistently poor families. Our data represent a sample of children in low income families, who have experienced poverty presumably for different lengths of time, and we have not concentrated on defining a persistently poor population. Undernutrition with the exception of calcium was not a concern in this population, but compared with NHANES II, there were 2.6 times as many overweight children.

In conclusion, obesity was the major concern in this low income multiethnic group of school children. Our data suggest caution in interpreting nutrient intake among overweight children because of an important underestimation of energy intake. Although the only micronutrient consistently low in this population was calcium in girls aged 10-12 y, intake of the more costly micronutrients is lower in children from poorer families or larger families. The expected association of dietary intake and energy expenditure was readily apparent in our study population. Because of the important health concerns of childhood obesity, current public health interventions in low income, ethnically diverse schoolchildren must focus on balancing energy intake and expenditure.