![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
School of Nutrition Science and Policy and the Jean Mayer U.S. Department of Agriculture Human Nutrition Research Center on Aging (HNRCA), Tufts University, Boston, MA 02111 and * School of Dietetics and Human Nutrition, McGill University, Montreal, Canada H9X 3V9
3To whom correspondence should be addressed. E-mail: tucker{at}hnrc.tufts.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
KEY WORDS: • dietary intake • dietary methodology • pregnancy • Africa • variance ratios • humans
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
,2
). In fact, most studies that have examined the sources of variation in dietary intake in developing countries have found lower intraindividual to interindividual variation ratios than those found in developed countries (3
,4
). However, subsistence communities may challenge this premise.
High variance ratios may result either from low between-person variability or, more often, from high intake variation within individuals. Statistically, within-person variance is the residual unexplained variance after all known systematic effects have been taken into account. It includes both true day-to-day variation in intake and measurement error (5
). Seasonal differences in food availability and effects from differing stage of pregnancy have been shown to contribute substantially to intraindividual variation in nutrient intakes in individuals from developing countries (2
,4
–9
). Additionally, individuals with low socioeconomic status (SES)4
may have higher intraindividual variability because of irregular food access. If, for example, animal products are consumed infrequently, then protein and/or fat intake may be high on a few days and low on most others. The effect of high intraindividual variation in intake on the variance ratio, however, also depends on the between-subject variation. If economic resources are severely limited and the link between food intake and income is strong, even small differences in income may lead to high interindividual variation. An understanding of the sources of variability in energy and nutrient intake in low income and subsistence communities is essential for determining the number of days of dietary intake per person that is required to accurately estimate individual usual intake.
Reliable estimates of energy and nutrient intakes are required if one is to examine associations between diet during pregnancy and maternal and child health (4
,5
,9
). Misclassification of individual intakes may substantially distort correlation coefficients, regression coefficients and relative risks if only a few days of replicate measures are taken (1
,4
,9
–11
). However, few studies have been conducted to assess the intraindividual and interindividual variation in energy and micronutrient intakes among pregnant subsistence farmers in developing countries. We hypothesized that due to the large seasonal variation, intraindividual variability in energy and micronutrient intakes would be greater than normally expected from developing countries because intakes are greater during the harvest than during preharvest seasons. We also expected interindividual variation in energy and nutrient intakes to be less than in other studies because of similar SES and the lack of dietary diversity among subsistence women in Malawi.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The study was conducted from 1988 through 1991 in TA Khongoni area in Lilongwe district. This area is located 
80 km west of Lilongwe City in the central region of Malawi. As in the rest of rural areas in Malawi, >90% of the population are subsistence farmers. Maize is their staple food. Malawi has three main seasons: a cool dry season from April to July, a hot dry season from August to November, and a rainy season from December to early April. For the purpose of this analysis, seasons were defined as (December–March) = preharvest, (April –July) = harvest, and (August –November) = postharvest. The harvest season is associated with increased food availability and reduced physical activity and the preharvest season is the period in which most farming activities occur and when food intakes are at their lowest (12
). Little agricultural work is performed in the postharvest season, but some households begin to experience food shortages around this time (13
).
Study protocol and subject recruitment.
To identify study participants, lists of village chiefs and their subchiefs were obtained from the District Commissioner’s office in Lilongwe. Based on both size and logistical considerations, the TA Khongoni area was selected. Village lists were used to conduct a population census within this area, and all women of child-bearing age were identified. From those identified, pregnant women were invited to join the study. After full explanation of the study procedures, verbal consent was obtained from all participants. The study was approved by the McGill University Human Ethics Review.
A total of 184 women with dietary observations during the 2nd and 3rd trimesters of pregnancy participated in the study. Most of the women were recruited into the study during the 2nd trimester of pregnancy and were followed until after completion of their pregnancy. Because women were contacted at various stages of pregnancy, the number of predelivery visits ranged from 1 to 3, with duration of 2–4 d each.
Selection and training of enumerators.
Enumerators for data collection were carefully interviewed and selected by the University of Malawi, Center for Social Research. All of those selected had a high school degree and spoke English and Chichewa fluently. The field supervisors had college degrees. Both enumerators and supervisors went through a 2-wk training program at Bunda College of Agriculture in Lilongwe in which the objectives of the study were explained and necessary skills in interviewing and observation were practiced. A nearby village served as a training site.
Dietary data collection.
Dietary intake data were collected using the weighed intake method. Enumerators lived in the villages in which they were working. During the day of the interview, they visited the study households beginning at 0600 h, when the women awakened, and stayed at the household until after the evening meal at 1800–2000 h when the women completed their day’s activities. They weighed the raw ingredients of all dishes before cooking, the final cooked dish, the subject’s portion and the remaining uneaten foods. In the earlier part of the study, dietary intake data were recorded on two consecutive days. This was extended to 4 d at each visit for women recruited later in the study. A total of 1061 diet days were collected, with a mean of 6 d/woman and total diet days ranging from 2 to 12. Plates and cups were supplied to all participants to assist with the individual food weighing. All foods were measured using Seca (Seca Corporation USA, Hanover, MD) and Salter scales (provided by UNICEF through the Ministry of Health, Malawi). Scales weighing 1 kg and 5 kg in divisions of 5 g were used for weighing small amounts of ingredients, individual potions, leftover or inedible food portions and snacks. Heavy family pots of cooked foods were measured using 10- to 25-kg scales. All liquid food items were measured in milliliters, using plastic graduated measuring cups (Shore Rubber Limited, Malawi). On the following day, women were asked to recall any late evening food or beverage consumption. All scales were calibrated regularly with known weights.
Data were entered into the MicroNAP Nutrient Analysis Program (14
) (Winnipeg, Canada). However, the nutrient file did not contain some of the foods that are consumed in Malawi such as white maize flour, certain vegetables, insects and termites. These foods were added into the database using the Composition of Foods Commonly Eaten in East Africa (15
) and Studies on the Chemical Composition of Foods Commonly Used in South Africa (16
).
Many calculations and variable interpretations were dependent on the stage of pregnancy. Although date of last menstrual period was asked, it was unreliable. Therefore, stage of pregnancy was estimated by subtracting each visit date from the birth date; and then dividing by 30 to determine month of pregnancy. Known premature deliveries were excluded from analysis.
Statistical methods.
Data analysis was done using SPSS for Windows, Version 10 (Chicago, IL). The distribution of each nutrient was tested for normality before analysis and skewed variables were log transformed. Means and standard deviations were analyzed using descriptive statistics. Mean differences by trimester were analyzed using independent sample t tests and one-way ANOVA was used to test mean differences by season. Variance components were analyzed with the Variance Components (VARCOMP) procedure in SPSS using the Restricted Maximum Likelihood Estimation method. Variance ratios were calculated as the error variance/variance across individuals. Within (intra) and between (inter) individual components were estimated by trimester and season of dietary intake. The within (CVw = Sw/mean of each nutrient) and between-individual (CVb = Sb/mean of each nutrient) CV were calculated using a formula provided by Beaton, (10
) where Sw and Sb are the square roots of the estimated intra- and interindividual variance obtained from the VARCOMP procedure in SPSS.
The CVw was then used to calculate the number of food weighing days to estimate true average intakes of individuals with the following formula from Beaton (10
):
 |
where n is the number of days needed per person, Z
is the normal deviate for the percentage of times the "true" average mean of the individual is expected to be covered by a confidence interval. For instance, Z = 1.96 for a 95% confidence interval. CVw is the intraindividual CV and Do is a specified confidence limit (as a percentage of true mean usual intake).
For example, if we wanted to calculate the number of days needed to estimate a person’s energy intake within 20% of their true mean 95% of the time, for a nutrient with CVw = 0.32, the calculation would be as follows:
 |
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
The mean age of the women was 26 ± 7.0 y, ranging from 14 to 51 y. Almost all (91%) of the women were married, and they had attended school for a mean of 2.8 ± 2.6 y. The mean years of schooling for their husbands was slightly higher than that of the women, at 4.7 ± 2.9 y. The average number of previous pregnancies per woman was four. Eighty-eight percent of the women had one or more previous pregnancies. Almost all of the respondents were participating in agricultural labor with 89% of the men and 92% of women reporting that their primary occupation was farming. Participants in the study areas lived in simple housing with all women living in houses having mud floors and grass thatched roofs, 76% lived in mud houses, whereas 19% had houses made of sun dried bricks (Table 1)
.
|
Intra/interindividual variations in energy and nutrient intakes in Malawian women ranged from 2.2 for energy to 12.7 for vitamin A. Examination of within and between CV (CVw and CVb, respectively) showed that high ratios resulted from low CVb for macronutrients except for fat and from high CVw for fat and micronutrients. Iron and vitamin A had both high CVw and low CVb, giving them the highest ratios (Table 2)
. Adjustment for season and pregnancy trimester did not improve these ratios. In fact, they increased for some nutrients.
|
) and Indonesian women (3
,4
) (Table 3)
. However, the variance ratios for U.S. nurses were similar (3.9 vs. 3.2) for protein and higher (11.7 vs. 5.3) for zinc. Ratios for Bangladesh women were higher than for Malawi women, but were available only for energy and protein (9
). Unpublished results from the Collaborative Research Support Program (CRSP) in Egypt, Kenya and Mexico lend further support for our results in Malawi. They found that variance ratios for energy, protein and fat were considerably higher in the two African countries, (ranging from 4.6–7 in Kenya and 4.9–10.1 in Egypt) than in Mexico (1.5–2.8). (Suzanne Murphy, Cancer Research Center of Hawaii, University of Hawaii at Manoa, personal communication).
|
–4
,9
). The intra/interindividual ratios for vitamin A were similar between pregnant women in Malawi and those in the United States, 12.7 and 11.7, respectively, whereas the ratio among Indonesian women was 3.2 (Table 3)
. Means and intra/interindividual variance components for energy and nutrient intakes by stage of pregnancy.
Mean energy intakes of the Malawian women were 7.3 and 6.9 MJ, and mean protein intakes were 57 and 54 g, during the 2nd and 3rd trimesters, respectively. Mean iron intakes were 13 mg in both trimesters. Energy and most nutrient intakes did not differ between the 2nd and 3rd trimesters (P > 0.05). However, women consumed more carbohydrates, fiber, calcium and vitamin C during the 2nd than 3rd trimester (P < 0.05). The intra/interindividual variance ratios for energy and all nutrients were generally lower during the 2nd trimester than the 3rd trimester, with the exception of iron for which the ratio during the 2nd trimester was greater than in the 3rd (9.1 vs. 8.1) (Table 4)
.
|
Energy intakes were higher during the harvest (7.5 MJ) and postharvest seasons (7.2 MJ) than in the preharvest season (6.5 MJ) (both P < 0.01). Fat, carbohydrates, protein, zinc, vitamins A and C and fiber intakes were also higher during the harvest and postharvest season than during the preharvest season (P < 0.05). Calcium and vitamin C intakes were greater during the preharvest than in the harvest and postharvest seasons (P < 0.05).
The variance ratios for energy, carbohydrates, fiber, and calcium, were greatest during the preharvest season. They ranged from 3.3 for energy to 5.4 for calcium. However, greater variance ratios for protein, fat, iron, zinc, vitamin A, vitamin B-12 and folate occurred during the postharvest season, ranging from 4.0 for protein to 99 for vitamin B-12 (Table 5)
.
|
The overall within-individual CV were lower for energy, carbohydrates, fiber and protein than for fat and micronutrient intakes. The number of dietary days required to estimate true individual average nutrient intakes within an error of ±10% is so large that it would not be feasible. If we accept an error range of ± 20%, 10 replicates would be needed to estimate energy, 8 to estimate carbohydrates, 21 to estimate fiber, 23 to estimate protein, 188 to estimate fat and 98 to estimate folate intakes. With an error range of ± 40%, two replicates would adequately estimate energy and carbohydrate intakes, 5 would be required for fiber, 6 for protein, 47 for fat and 16 (iron) to 53 (vitamin A) for micronutrients (Table 6)
.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
,4
) and Bangladesh (9
), that have published nutrient variance ratios for women in developing countries. Additional unpublished variance ratios have been reported for Kenyan, Egyptian and Mexican women. We found higher intra/interindividual variance ratios in Malawian subsistence farming women than those found in developed countries (2
,10
,11
,17
). Although the studies in Bangladesh (9
) and the relatively high variance ratios in Kenyan and Egyptian women (Murphy, personal communication) support our findings in developing countries, two studies from Indonesia found that variance ratios for energy and carbohydrates were <1 (CVb was greater than CVw) (3
,4
), suggesting great variation across developing countries. In rural areas of Malawi, foods from animal sources are rarely eaten. Most of the energy intake comes from grains and cereals, and most iron from vegetable sources. It is likely that this occasional and irregular use of animal foods contributes greatly to the high variance ratios seen. Explanations for the high within individual variance ratios in Malawian subsistence farmers include the dependence of the population on the overall and seasonal availability of locally produced food. Although not representative of Malawi as a whole, we believe our results reflect the true intra/interindividual variation in nutrient intakes in the population of rural subsistence farmers.
As with previous studies, we found that variance ratios for micronutrients were generally greater than for energy and macronutrients (1
,3
,4
,10
,11
). This suggests that it will be more difficult to find associations between micronutrient intake and maternal health outcomes than with energy and macronutrient intakes in this population because micronutrients are measured with greater error than energy and macronutrients (4
). The number of replicate days required to estimate the mean intake of individuals varied from nutrient to nutrient. Persson (4
) found that six replicates would be required to estimate true individual intakes with precision of ± 20% for energy, carbohydrates, vitamin A, iron and vitamin C in Indonesia. One of the reasons for the lower intra- to interindividual variance ratios in the Indonesian studies could be that in some developing countries in which a limited number of foods are consumed and in which food consumption is more closely associated with income than with local availability, there is greater person-to-person, relative to within-person variation (4
).
In the pregnant subsistence farmers of this study, energy, carbohydrates, fiber and protein were the only nutrients that could be estimated reasonably (10 replicate d) within an error range of ± 30%. Other nutrients would require from 29 to 95 d at this level of error. However, the number of days required to estimate intake depends on the intended use of the data. Use for group measures or some comparative analyses, for example, may not require the same degree of precision as does estimation of individual usual intake (5
). Because the lowest intra/interindividual variation ratios occurred during the harvest season and highest ratios during the preharvest season, we concluded that increased variability resulted from greater food shortages during the preharvest season, leading to greater intraindividual variability because food availability fluctuates. During the harvest season, when most women had access to food, intraindividual variation was much lower, which continued to support the observation that seasonal food availability significantly affected estimates of usual intake.
In conclusion, despite limited dietary diversity in this subsistence-based population, large within-person variation in nutrient intake poses a challenge for dietary assessment in studies requiring estimates of usual intake when there may be large seasonal variation in food supply. More, rather than fewer days of dietary intake were required to correctly rank usual intakes of pregnant subsistence farmers relative to Western populations and to more urbanized developing country populations.
Variance ratios tended to be greater in the 3rd than in the 2nd trimesters. We also noted that dietary intakes tended to be higher for carbohydrates, fiber, calcium and vitamin C during the 2nd than the 3rd trimester. However, it was seasonal availability in food and not pregnancy per se that contributed to the higher intraindividual variation in nutrients in the pregnant subsistence farmers. Thus the contrast in variance ratios in Malawian women relative to those from Indonesia (3
,4
) suggests that results from one country cannot be generalized to others. It is likely that these ratios also differ within countries by level of urbanization, SES and other factors that affect food availability and intake. Within this population, the finding that variance ratios vary by season, in addition to overall seasonal difference in intake, underscores the importance of considering these issues for each nutrient of interest during study design.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Supported through The International Center for Research on Women and funded by Cooperative Agreement #DAN-1010-A-00–7061-00 with the Offices of Nutrition and Health of the U.S. Agency for International Development, the International Development Research Centre of Canada (IDRC), agreement #88–0342, and an Équipe (Team) grant from "Fonds de Recherche en Santé de Québec" (FRSQ), a Quebec provincial funding agency.
4 Abbreviations used: CVb, interindividual coefficient of variation; CVw, intraindividual coefficient of variation; Do, the specified confidence limit (as a percentage of long-term true mean intake); SES, socio-economic status; Sb, square root of the estimated interindividual variation; Sw, square root of the estimated intraindividual variation; VARCOMP, variance components; Z, the normal deviate for the percentage of times the "true" average mean of the individual is expected to be covered by a confidence interval.
Manuscript received 31 October 2991. Initial review completed 7 January 2002. Revision accepted 23 February 2002.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
1.
Tarasuk, V. & Beaton, G. H. (1992) Statistical estimation of dietary parameters: implications of patterns in within-subject variation—a case study of sampling strategies. Am. J. Clin. Nutr. 55:22-27.
2. Willett, W. (1998) Nutritional Epidemiology 2nd ed. 1998 Oxford University Press Oxford, UK. .
3. Launer, L. J., Kardjati, S., Kusin, J. A. & Reed, G. F. (1991) Patterns of variability in the nutrient intake of nutritionally vulnerable pregnant women. Eur. J. Clin. Nutr. 45:131-138.[Medline]
4.
Persson, V., Winkvist, A., Hartini, T. N. S., Greiner, T., Hakimi, M. & Stenlund, H. (2001) Variability in nutrient intakes among pregnant women in Indonesia: implications for the design of epidemiological studies using the 24-h recall method. J. Nutr. 131:325-330.
5.
Beaton, G. H. (1994) Approaches to analysis of dietary data: relationship between planned analyses and choice of methodology. Am. J. Clin. Nutr. 59:253S-261S.
6. FAO/WHO/UNU (1985) Energy and Protein Requirements. World Health Organization Technical Report Series No. 724 1985 WHO Geneva, Switzerland. .
7. Schaefer, A. E. (1981) Can nutritional status be determined from consumption or other measures?. National Academy of Sciences: Assessing Changing Food Consumption Patterns 1981 National Academy Press Washington, DC. .
8.
Brown, K. H., Black, R. E. & Becker, S. (1982) Seasonal changes in nutritional status and prevalence of malnutrition in a longitudinal study of young children in rural Bangladesh. Am. J. Clin. Nutr. 36:303-313.
9. Torres, A., Willett, W., Orav, J., Chen, L. & Huq, E. (1990) Variability of total energy and protein intake in rural Bangladesh: implications for epidemiological studies of diet in developing countries. Food Nutr. Bull. 12:220-228.
10.
Beaton, G. H., Milner, J., Corey, P., McGuire, V., Cousins, M., Stewart, E., de Ramos, M., Hewitt, D., Grambsch, V., Kassim, N. & Little, J. A. (1979) Sources of variance in 24-hour recall data: implications for nutrition study design and interpretations. Am. J. Clin. Nutr. 32:2546-2559.
11.
Sempos, C., , T., Johnson, N. E., Smith, E., , L & Gillian, C. (1985) Effects of intraindividual and interindividual variation in repeated dietary records. Am. J. Epidemiol. 121:120-130.
12. Ndekha, M., Kulmala, T., Vaahtera, M., Cullinan, T., Salin, M.-L. & Ashorn, P. (2000) Seasonal variation in the dietary sources of energy for pregnant women in Lungwena, rural Malawi. Ecol. Food Nutr. 38:605-622.
13. Government of Malawi/United Nations (1993) Malawi: situation analysis of poverty 1993 Government of Malawi and United Nations in Malawi Lilongwe. .
14. Northern Technical Data (1989) MicroNap: Nutrient Analysis for Research Purposes, Version 4.21 1989 Northern Technical Data Winnipeg, Canada. .
15. West, C., Pepping, F. & Temalilwa, C. R. (1988) The Composition of Foods Commonly Eaten in East Africa 1988 Wageningen Agricultural University Wageningen, the Netherlands. .
16. Fox, F. W. (1986) Studies on the Chemical Composition of Foods Commonly Used in Southern Africa 2nd ed. 1986 South African Institute for Medical Research Johannesburg, S. Africa. .
17. Egami, I., Wakai, K., Kaitoh, K., Kawamura, T., Tamakoshi, A., Lin, Y., Nakayama, T., Sugimoto, K. & Ohno, Y. (1999) Intra- and inter-individual variations in diets of the middle-aged and elderly Japanese. Jpn J. Public Health 46:838-837.
This article has been cited by other articles:
|
|
 |
|
  I. Huybrechts, D. De Bacquer, B. Cox, E. H. Temme, H. Van Oyen, G. De Backer, and S. De Henauw Variation in energy and nutrient intakes among pre-school children: implications for study design Eur J Public Health, June 4, 2008; (2008) ckn017v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Jahns, A. Carriquiry, L. Arab, T. A. Mroz, and B. M. Popkin Within- and Between-Person Variation in Nutrient Intakes of Russian and U.S. Children Differs by Sex and Age J. Nutr., November 1, 2004; 134(11): 3114 - 3120. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. P. Murphy, C. Gewa, L.-J. Liang, M. Grillenberger, N. O. Bwibo, and C. G. Neumann School Snacks Containing Animal Source Foods Improve Dietary Quality for Children in Rural Kenya J. Nutr., November 1, 2003; 133(11): 3950S - 3956. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
