点击显示 收起
1 From the Department of Preventive Medicine, Belgium; the Department of Medicine, University of Göteborg, Göteborg, Sweden; and the Department of Internal Medicine, Division of Diabetes and Nutrition, Catholic University of Louvain, Brussels.
2 This work is part of a PhD thesis at the Department of Public Health, Catholic University of Louvain, and the Unit of Pediatric Gastroenterology and Nutrition, University Clinics St Luc, Brussels. 3 Address reprint requests to M Guillaume, Department of Preventive Medicine, Province de Luxembourg, Observatoire de la Santé, Rue de la Station 49, 6900 Marloie, Belgium. E-mail: obs.sante{at}province.luxembourg.be.
ABSTRACT
Background: Sex differences in the effects of genetic and environmental factors on circulating lipids have been examined mainly in adults, in whom the influences of sex steroid hormones are well known.
Objective: Our objective was to determine the effect of sex on genetic and environmental influences on serum lipids in prepubertal boys and girls.
Design: Children aged 6–8, 8–10, and 10–12 y (n = 1028) were selected at random in the Belgian province of Luxembourg, a region in Europe with a high prevalence of risk factors for cardiovascular disease and diabetes. Blood glucose and serum cholesterol, triacylglycerol, and insulin concentrations were measured, and anthropometric data and blood pressure were recorded. Familial data were obtained from standardized questionnaires. Nutritional status was obtained from a 3-d record. Participation was 70.3% of the primary cohort.
Results: Cholesterol, triacylglycerol, and insulin values were among the highest recorded in studies of children. In girls, cholesterol correlated positively with the energy density of intake of saturated fat (r = 0.13, P = 0.001), cholesterol (r = 0.11, P = 0.006), and protein (r = 0.12, P = 0.007) and negatively with the ratio of polyunsaturated to saturated fat intake (r = -0.14, P = 0.001) and the energy density of carbohydrate intake (r = -0.11, P = 0.019). In boys, no such relations were found. Triacylglycerol was not significantly related to nutritional factors. Consistent, independent relations were found between reported elevated cholesterol concentrations in the parental and grandparental generation and cholesterol (r = 0.101, P = 0.011) and triacylglycerol (r = 0.09, P = 0.03) in boys. No such associations were found in girls.
Conclusion: Environmental and genetic factors may have different effects on serum cholesterol in girls and boys.
Key Words: Children • cholesterol • triacylglycerol • nutrition • familial factors
INTRODUCTION
The production and turnover of circulating lipoproteins are clearly influenced by nutrition. Both total energy intake and individual macronutrient intake interact with plasma lipid concentrations. Of particular importance is the quantity and quality of fat intake, as well as that of cholesterol and, in some instances, carbohydrate. These associations have been studied frequently in adults, whereas reports in children are less numerous (1).
There are multiple genetic influences on lipoprotein metabolism, from production to breakdown or storage via receptors and enzymatic pathways (2). Again, these influences have been studied mainly in adults. However, such studies would presumably be more rewarding in children, in whom genetic factors should be easier to detect before the secretion of sex steroid hormones, which exert profound influences on lipoprotein metabolism.
The Luxembourg province in the southern part of Belgium is an area with a known high prevalence of dyslipidemia, obesity, type 2 diabetes, and cardiovascular disease among adults (3). In 1992, cohorts of boys and girls of different ages were selected at random and anthropometric and blood chemistry measurements were analyzed in relation to familial, socioeconomic, psychosocial, and lifestyle factors. In addition, a 3-d record of nutritional habits was collected. We report here the statistical associations between nutritional factors in the children and their familial predisposition to hypercholesterolemia on one hand and their blood glucose and serum concentrations of lipids and insulin on the other.
SUBJECTS AND METHODS
Population
The Luxembourg province of Belgium (Province de Luxembourg) is located in the southern, French-speaking (Wallonian) region of Belgium. The province borders Luxembourg on the east and France on the south and southwest. The province has a low population density compared with the rest of Belgium (4) and a small proportion of non-Belgian citizens (4). Immigration has been low and the population is relatively stable (4). The area is characterized by nonindustrial occupations, such as public service activities, and is mainly rural, particularly in the southern part. The mean per capita income is less than that in other parts of Belgium and the mean educational level is also comparably lower (5).
The selection of the studied cohorts was described in detail previously (6). In short, school classes with children aged 6–8, 8–10, and 10–12 y were selected with use of a randomization procedure, taking into consideration the different school systems in Belgium (consisting of communal, state, and private schools). In total, 1462 children were selected in the primary cohort and 1028 (70.3%) participated in the study. As assessed by examining school records, nonparticipants did not differ significantly from participants in anthropometric variables but were apparently from a lower social class (6). The examinations were approved by the Ethical Committee of the Catholic University of Louvain, Brussels; the Departments of Health, Welfare, and Education of the Belgian government; and the political authorities of the Province de Luxembourg.
Methods
The children were examined in 1992. Anthropometric and blood chemistry data were collected and questionnaires covering familial data and environmental factors were administered. These methods and the quality controls were described in detail previously (6) and are therefore only summarized briefly here.
For the measurement of blood chemistry variable, capillary blood was obtained in the morning after the children had fasted overnight. The anthropometric data collected included height, weight, waist and hip circumferences, and subscapular and triceps skinfold thicknesses. From these measurements, body mass index (BMI; in kg/m2) and waist-to-hip ratio (WHR) were calculated. Questionnaires included information on familial aggregation of disease, including elevated cholesterol; lifestyle; physical activity; demographic data; birth data; and social and familial conditions.
Familial data were obtained from a questionnaire self-administered by the parents. Recorded answers were cross-checked during interviews conducted in the subjects' homes. Information on reported elevated cholesterol values was recorded in 5 categories: 1) absence of problems (n = 218), 2) a positive event reported in 1 parent only (n = 17), 3) a positive event reported in 1 grandparent only (n = 242), 4) a positive event reported in 1 parent and 1 grandparent (n = 51), and 5) a positive event reported in 1 parent or 1 grandparent (n = 431). Lack of knowledge was reported in 386 cases.
Nutrient intake was estimated from 3-d records for 955 children. The days selected were 2 consecutive weekdays and 1 weekend day (7). The children, assisted by their teachers and parents, were responsible for the recordings. The children were carefully instructed how to use their food consumption records and how to estimate portion sizes. Teachers and parents obtained written information.
As soon as possible after the recording, an interview was planned at each subject's home, preferably in the presence of the mother or the person in charge of preparing meals if this was not the mother. During this interview, the reliability of the record was checked, and additional information was collected to further improve the accuracy of the recording of the quality and quantity of food intake (eg, the kind of food items, details about recipes, amount of sugar or salt added, and description of quantities eaten). To test the reliability of the record, several entries were crosschecked during the interview. When the children had chosen to eat at school, the menus served were recorded.
The quantitative aspect of food intake was estimated by using household measurements that were afterward converted into grams of food or milliliters of beverages with use of an equivalence table. The estimates made by the different observers during a training session preceding the actual study did not differ significantly. After the data were encoded, a food-composition table was used to estimate the amount of nutrients consumed. The food-composition table used was based on that of the MONICA project in Belgium (8, 9) with some new food items.
Data processing and statistical methods
Data were stored and analyzed in an IBM computer (IBM, Armonk, NY). Different procedures from SAS were used to analyze the data (10). Analysis of variance (ANOVA), with the GLM procedure, was applied for comparisons of means between groups. Pairwise comparisons were made with Student's t test. For both sexes, the relations between continuous variables were studied by using stepwise multiple regression and correlation (REG and CORR procedures).
Nutritional data appeared to be normally distributed. For the variables apparently deviating most strongly from the assumption of normal distribution (triacylglycerol and insulin), results were retested after logarithmic transformation or by use of a nonparametric test (Wilcoxon). No significant differences were found in the results obtained (not shown). Adjustments were made for spurious significances due to multiple testing.
Data were adjusted for age when indicated and all analyses were performed separately for boys and girls. Two-tailed tests were used and P values <0.05 were considered to indicate significance.
RESULTS
Mean values of cholesterol, triacylglycerol, glucose, and insulin in boys and girls in the different age groups are shown in Table 1. About 90% of the children gave complete answers to the nutritional survey; the nutritional data are summarized in Table 2. The data in Tables 1 and 2 were partly reported previously (6, 7) but are included here for convenience.
View this table:
TABLE 1.. Metabolic variables in boys and girls1
View this table:
TABLE 2.. Nutritional intake in boys and girls
Correlations were performed between nutritional factors and serum lipid, insulin, and blood glucose values (Table 3). In boys, no significant relations with nutritional data were found for cholesterol. In girls, cholesterol concentrations correlated directly and significantly with intake of total energy, intake of total lipids (borderline), and the energy density of saturated fat, cholesterol, and protein intakes. Significant negative relations were found in girls with the ratio of polyunsaturated to saturated fat in the diet (P:S) and the energy density of carbohydrate intake. These relations remained significant after adjustment for the influence of triacylglycerol, insulin and glucose, blood pressure, and BMI.
View this table:
TABLE 3.. Correlations between cholesterol and insulin concentrations and nutritional factors expressed in absolute values and as energy density in boys and girls1
In boys, insulin showed significant relations with the energy density of saturated fat and protein intakes (positive), the P:S (borderline), and the energy density of carbohydrate intake (negative). In girls, insulin showed borderline significant correlations with total energy consumption and the energy density of lipid intake and negative correlations with the energy density of carbohydrate and fiber intakes. Except for the energy density of fiber, these correlations became insignificant after multiple regression analyses that took lipids, glucose, BMI, and blood pressure into account. Triacylglycerol values were not correlated with any nutrient (not shown) and blood glucose was negatively correlated with the energy density of fiber intake (P = 0.026).
The familial influence on serum lipid values was analyzed by taking different categories of familial aggregations into account. Comparisons were performed with ANOVA after the data were adjusted for age. In girls, no significant associations with familial reports of elevated cholesterol were found. In boys, reported elevated cholesterol in the family at level 4 (parents and grandparents) was directly associated with the cholesterol values of the boys (Figure 1; P = 0.011, n = 22), as was reported elevated cholesterol in the family at level 5 (parents or grandparents; P = 0.057, n = 215). The cholesterol relations at level 4 remained significant in multiple regression analyses after adjustments for BMI, triacylglycerol, insulin, blood glucose, and blood pressure (P = 0.011). In further analyses that took age, physical activity, birth factors, nutrition, and the socioeconomic status of the father into account (in our model, the socioeconomic status of the father was highly correlated with the socioeconomic status of the family), reported elevated cholesterol at level 4 was the only remaining variable having a significant association with cholesterol values in boys, explaining 7.8% of the variance.
FIGURE 1. . Mean (±SD) total serum cholesterol concentrations in boys with () or without () a family history of elevated cholesterol concentrations in their parents or grandparents. Values were calculated for each age in boys with or without a family history. For statistical comparison, the values were first age-adjusted and then compared by using ANOVA. Age adjusted difference, P = 0.011.
Furthermore, in boys, triacylglycerol values were significantly correlated with familial reports of elevated cholesterol in parents (level 2; P = 0.006, n = 6), in grandparents (level 3, P = 0.028, n = 122), and in parents or grandparents (level 5; P = 0.025, n = 215) (Figure 2). These relations remained significant in multiple regression analyses after adjustment for BMI, cholesterol, insulin, blood glucose, blood pressure, and the socioeconomic status of the father (P = 0.004, 0.045, and 0.044, respectively), explaining 9.7% of the variance. Age, physical activity, and birth weight did not contribute to the explanation.
FIGURE 2. . Mean (±SD) triacylglycerol concentrations in boys with () or without () a family history of elevated cholesterol concentrations in their parents or grandparents. Statistical treatment was as in Figure 1. Age-adjusted difference, P = 0.025.
DISCUSSION
The children examined in this study had among the highest serum lipid concentrations reported in children of comparable ages, as discussed previously (6). We therefore considered it of interest to analyze the associations between these measurements and potential environmental and familial interactions.
The participation rate for the present study was 70.3% of the primary cohort. Some data were available for nonparticipants from registers, telephone contacts, and school records. Nonparticipating children were not significantly different from participants in weight or height, but their families had lower socioeconomic statuses (6). The effect of this on the results reported here might be primarily an influence on food habits because, among the participants, eating patterns characterized by high fat and low carbohydrate contents were inversely related to the father's education (7). Whether the familial clustering of problems with hypercholesterolemia was affected by a bias due to the absence of data from nonparticipants is not known. When these data are interpreted in terms of their generalizability, this should be kept in mind.
The selection of categories for the collection of familial data had the following rationale. Level 1 was devoid of reported problems. Level 2 examined relations between elevated cholesterol in parents only and the metabolic variables in their children, in other words, through one generation. Level 3 analyzed only grandparents and level 4 analyzed parents and grandparents. The latter was chosen in an attempt to find stronger familial relations, registered through 2 generations. Level 5 recorded events in parents or grandparents to obtain information from as many family members as possible to increase the number of observations.
We found associations between cholesterol and nutritional data only in girls. In girls, the positive relations between cholesterol concentrations and high intakes of total and saturated fat, cholesterol, and protein and a low P:S and low carbohydrate intake were independent of other measured variables, including obesity, which by itself had a limited effect on cholesterol concentrations in this cohort (6). Such a nutritional pattern would be expected to be followed by elevated cholesterol concentrations (1).
There were several relations between nutritional factors and insulin, including direct associations with intakes of energy, fat, and protein and an indirect association with carbohydrate intake. These variables are also correlated with obesity, which was prevalent in this cohort of children (6, 11). When the relations were adjusted for BMI, these significant relations disappeared. We therefore suggest that the association between insulin and nutritional status was mediated mainly via obesity.
Another environmental factor that might be involved in regulating circulating lipid concentrations is physical activity (12). Neither involvement in sports nor the amount of time or frequency of watching television (as indexes of physical activity and inactivity, respectively), however, was related to cholesterol or triacylglycerol concentrations (13). This observation suggests that physical activity was not an important determinant of lipid concentrations.
Taken together, these data indicate that nutritional status is of more importance for the determination of cholesterol concentrations in girls than in boys. In contrast, however, familial relations with cholesterol were found only in boys. In particular, the independent relation between reported elevated cholesterol in both the parents and the grandparents and cholesterol values in the boys suggests a strong familial aggregation of elevated cholesterol. Similarly, triacylglycerol concentrations, in boys only, were independently related to elevated cholesterol in 2 generations and this relation was independent of obesity, a strong determinant of triacylglycerol in this cohort of children (7).
A total of 37.5% of the participants reported a lack of knowledge of familial information on blood lipids. Among these participants might be those with families with no lipid problems or with unknown problems, including those in whom lipid concentrations had not been measured. The fraction of children with such responses from their parents were compared with the rest of the population and with those who reported no familial lipid problems and no significant differences were found in cholesterol, triacylglycerol, glucose, insulin, blood pressure, BMI, nutritional data, or socioeconomic variables (not shown), suggesting no selection bias. Most cholesterol measurements had been performed in yearly checkups or surveys of the families, but some probably were made as a result of disease associated with elevated cholesterol concentrations. The latter probably introduced a nonrandom enrichment of the data with elevated cholesterol values. There is also a possibility that the category of "no problem" was underestimated, because it seems likely that elevated lipid values would be better known between family members than would normal values. It seems likely that both these problems were more pronounced in the grandparental than in the parental generation, because the parents provided the information.
A main finding of this study was the striking difference in the familial effect on lipid concentrations between boys and girls. It seems unlikely that this observation might be invalidated by the selection bias discussed, because such bias should have had an equal effect on boys and girls.
Family members often share food habits (14). It might therefore be considered that the familial associations found were determined by similar eating habits among family members. This is unlikely, however, because if this were the case, both boys and girls would have shown similar relations to familial data. Furthermore, the grandparents, with whom familial associations were also found, usually lived in separate households and therefore did not eat the same meals as did the children and their parents.
Taken together, these results indicate that in girls nutritional status was of importance for the association with cholesterol values, but not for triacylglycerol values, whereas no such relations were found in boys. In contrast, in boys, the results indicate familial aggregation for both cholesterol and triacylglycerol values, which was not found in girls. This strong familial association was also clearly found in boys after adjustment for potential confounders. These results suggest that environmental and genetic influences may be different in boys and girls.
Several genetic abnormalities of lipoprotein metabolism have been disclosed in adults, with no apparent evidence of sex specificity (2). It seems possible that in adults sex differences are hidden by the marked influence of sex steroid hormones on lipoprotein metabolism. If this is correct, then differences between the sexes in the expression of genetic dyslipidemia would be easier to disclose in children before puberty, such as in the present study. The type of potential abnormality present can only be speculated on in the absence of detailed lipoprotein data.
In summary, environmental effects on cholesterol concentrations (in terms of nutrition) were found in girls only, whereas familial hypercholesterolemia was associated with both cholesterol and triacylglycerol in boys only, suggesting a different effect of nutritional and familial factors on lipid concentrations in boys and girls. Whether the lack of association between triacylglycerol and nutrition in girls means that such factors affect mainly cholesterol metabolism, or whether genetic factors were expressing a mixed hyperlipidemia in boys, has yet to be clarified.
REFERENCES