Vitamin A deficiency (VAD), even subclinical vitamin A deficiency (SVAD), could increase the risks of morbidity and mortality in childrenReference McLaren and Frigg1, 2. At least 100 million of the world's under-fives are vitamin A deficientReference McLaren and Frigg1. Breast milk, vitamin A-rich foods and foods fortified with vitamin A are recommended and used for prevention of VADReference Sommer and West3, 4. Promoting breast-feeding is regarded as the best way to protect babies from VAD in public health campaignsReference Sommer and West3, Reference West, Chirambo, Katz and Sommer5. However, several studies reported a high prevalence of VAD in breast-fed childrenReference Ettyang, Oloo, van Marken Lichtenbelt and Saris6, Reference Oso, Abiodun, Omotade and Oyewole7. In some developing countries, VAD has been found to be even more common among women than in childrenReference Katz, Khatry, West, Humphrey, Leclerg and Kimbrough8, Reference West9, and a deficient mother might not give adequate vitamin A to her child because of low concentrations of vitamin A in her breast milkReference Ettyang, Oloo, van Marken Lichtenbelt and Saris6, Reference Katz, Khatry, West, Humphrey, Leclerg and Kimbrough8, Reference Gross, Hansel, Schultink, Shrimpton, Matulessi and Gross10 Furthermore, a study reported that lactating mothers, especially those using prolonged breast-feeding, would be less likely to supply complementary foods adequately or in a timely manner for their childrenReference Fawzi, Herrera, Nestel, el Amin and Mohammed11, and breast milk alone may not meet the entire needs for a child aged >6 months. Most of the studies that have reported increased risks of SVAD in breast-fed children were from countries or regions where there was severe VAD. Associations of VAD of children with breast-feeding and dietary intakes have not been well characterised in countries where there is mild VAD, for example in China. The purpose of this study was to report the prevalence of SVAD and explore the associations between SVAD and breast-feeding, complementary food intakes and other risk factors in children aged 0–5 years in Anhui Province, China.
The cross-sectional study was conducted in December 2002 in Anhui Province, one of the agricultural provinces in the middle east of China, with a population of >60 million. A total of 1097 children aged between 0 and 5 years were recruited for the study using the proportionate-to-population size cluster sampling method from four urban districts in one city and 24 rural villages in four counties. If more than one eligible child lived in a household, investigators randomly selected one for the survey.
Mothers or caregivers were asked to bring the children selected for the survey to the local township hospitals for evaluation. After verbal consent was obtained from children's parents or caregivers, mother or caregivers were interviewed about the child's demographics, breast-feeding and other food consumption in the past week, any vitamin A supplement in the past 6 months and the presence of diseases during the previous 2 weeks. Information on the mother's their demographics, reproductive history and family's socio-economic status was also collected.
Questions about breast-feeding practices were asked: ‘Has your child been breastfed in the past week? If “Yes” what type of breast-feeding? Exclusive or non-exclusive breast-feeding?’ Children who were breast-fed, and with no supplementation of any type (no water, no juice, no non-human milk and no foods) except for vitamins, minerals and medications, were categorised into the exclusive breast-feeding group12; the remainder were classified into the non-exclusive breast-feeding group.
Food consumption in the past week was also assessed using a semi-quantitative food-frequency questionnaire, which was designed to include all important food sources of vitamin A and carotenoids in the typical local diet and used in the national survey for VAD in China in 2001Reference Lin, Liu, Ma, Tan, Zhang and Jiang13. The questions were ‘In the past week, how often did your children eat (or drink) any of following foods: cow's milk or its products, cereals, eggs, pulses or soy foods, red and yellow fruits, dark green leafy vegetables, fish and shrimps, meat and animal liver?’ Response categories for every type of foods were none, once or more a week and at least once a day.
Vitamin A supplement intake was determined using the question: ‘Has your child had any supplementation of fish-liver oil or other vitamin A preparations in the past 6 months?’
About 0.2 ml of capillary blood was drawn by finger stick for serum retinol determination from all participants in a dark room and placed in a plastic EDTA tube without anticoagulant. Serum was obtained by centrifugation within 30 min. The serum tube was protected from light by wrapping it in aluminium foil, placed in a cold chain box containing packets of dry ice and then frozen at − 20°C within 4 h. All specimens were finally transported on dry ice to the laboratory of the Capital Institute of Pediatrics and stored at − 70°C until assay. Serum retinol was detected at an emission wavelength of 480 nm and an excitation wavelength of 330 nm with a Shimadzu RF-540 fluorescence detector. The analytical protocols and laboratory quality assurance procedures of the US National Health and Nutrition Examination Survey (NHANES)Reference Lin, Liu, Ma, Tan, Zhang and Jiang13 were followed by this laboratory.
The cut-off value for SVAD was defined as serum retinol ≤ 20 μg dl− 1 according to criteria recommended by the World Health Organization. This research protocol was reviewed and approved by the Ethical Committee in Anhui Medical University.
We used the χ2 test for categorical variable analysis and the Cochrane–Armitage method for χ2 trend analysis. To control confounding as well as to test the possible interactions between breast-feeding and other factors, we estimated and tested the odds ratios in the logistic model with interaction using dummy coding14. We conducted all the data analysis using SAS software version 8.2 (SAS Institute Inc.). All tests were two-sided and were considered significant at 0.05.
Characteristics of the study population
A total of 1097 children aged 0–71 months were recruited for this study. Forty-five children were excluded due to lack of vitamin A data. Of 1052 children, 72 SVAD cases were found. The prevalence rate of SVAD was 6.9% in all children, and 13.0, 9.1, 5.3, 2.9, 4.3 and 4.7% in children aged 0, 1, 2, 3, 4 and 5 years, respectively.
The breast-feeding rates were 78.33% (159/263), 23.74% (52/219) and 5.12% (11/215) for children aged 0, 1 and 2 years, respectively. The breast-feeding rates declined with age (Cochrane–Armitage trend test, P < 0.01).
Table 1 shows the characteristics of children with SVAD and non-SVAD. Gender, fever in the previous 2 weeks, and cold, cough or other infection of the respiratory tract in the previous 2 weeks were similar between the two groups.
* χ2 test, P < 0.05.
** χ2 test, P < 0.001.
† The cut-off value for SVAD was defined as serum retinol ≤ 20 μg dl− 1.
Significant differences between SVAD and non-SVAD groups were found in age in categories, place of residence, rank among siblings, mother's education, mother's occupation, vitamin A preparation supplementations, and diarrhoea in the previous 2 weeks.
The odds ratios of SVAD for breast-feeding in all children and within age categories
The odds ratio of SVAD for breast-feeding was 3.56 (95% confidence limits (95% CL) 2.17–5.82). After adjustment for sex, age in categories, residence, mother's education, mother's occupation, vitamin A preparation supplements, rank in siblings and diarrhoea, the odds ratio for breastfeeding fell to 2.38 (95% CL 1.13–4.95), but remained statically significant (Table 2).
OR – odds ratio; 95% CL – 95% confidence limits.
** P < 0.01.
† The cut-off value for SVAD was defined as serum retinol ≤ 20 μg dl− 1.
‡ Adjusted for breast-feeding, sex, age in categories, rank in siblings, mother's education, mother's occupation, vitamin A preparation supplement, and diarrhoea in the previous 2 weeks.
§ Breast-feeding: children breast-fed in the past week were categorised as breast-feeding, otherwise as non-breast-feeding. For children in the breast-feeding category, those with no supplementation of any type (no water, no juice, no non-human milk and no foods) except for vitamins, minerals and medications were categorised in the exclusive breast-feeding group, otherwise in the non-exclusive breast-feeding group.
¶ Odds ratios were estimated and tested in the logistic model with interaction using dummy coding.
We estimated the odds ratios for breast-feeding within age categories using logistic regression with interaction by dummy coding. The odds ratios for breast-feeding did not show statistical significance within age 0 and 2 categories. The odds ratios for breast-feeding within age 1 category were 5.46 (95% CL 2.07–15.03) and 4.6 (95% CL 1.72–12.82) before and after adjustment for other confounders, respectively, and both were statistically significant (Table 2).
The odds ratios for exclusive breast-feeding did not show statistical significance within age 0, 1 and 2 categories, although the odds ratios for exclusive breast-feeding reached 6.29 and 8.93 before and after adjustment for other confounders, respectively, within age 1 category. The odds ratios for non-exclusive breast-feeding within age 1 category were 5.39 (95% CL 1.99–15.1) and 4.36 (95% CL 1.58–12.01), and showed statistical significance before and after adjustment of other confounders, but no statistical significance was found for that within the age 0 and 2 categories (Table 2).
The odds ratios for breast-feeding decreased, but remained statically significant, after further adjustments for individual dietary factors or all dietary factors (data not shown).
The effects of dietary intakes on SVAD
Cochrane–Armitage χ2 trend tests showed negative or decreasing trends between the SVAD and the frequencies of food intakes, except for meat and liver. Univariate logistic regression analysis showed that the intakes of milk or its products, eggs, pulses or soy foods, red and yellow fruits, dark green vegetables, fish and shrimps, etc. had protective effects on SVAD (Table 3). After adjustment for breast-feeding, sex, age in categories, rank in siblings, mother's education, mother's occupation, vitamin A preparation supplements and diarrhoea, no food intake was found to have a significant effect on SVAD (Table 3).
OR – odds ratio; 95% CL – 95% confidence limits.
** Cochrane–Armitage χ2 trend tests, P < 0.001.
† The cut-off value for SVAD was defined as serum retinol ≤ 20 μg dl− 1.
‡ Frequencies of intake of foods refers to the frequencies of food intakes in the past week, and response categories for every type of foods were none, once or more a week and at least once a day.
§ Adjusted for breast-feeding, sex, age in categories, rank in siblings, mother's education, mother's occupation, vitamin A preparation supplement, and diarrhoea in the previous 2 weeks.
The effects of other factors on SVAD
The crude odds ratios of SVAD for residence (4.25, 95% CL 1.53–11.79), mother's education (2.83, 95% CL 1.12–7.13), mother's occupation (2.94, 95% CL 1.44–5.99), breast-feeding, vitamin A preparation supplements (0.28, 95% CL 0.12–0.65), rank in siblings (2.49, 95% CL 1.51–4.12) and diarrhoea (2.24, 95% CL 1.22–4.1) were statistically significant. For those factors, only rank in siblings (odds ratio 2.08, 95% CL 1.79–3.62) remained statistically significant in the full adjusted model including sex, age in categories, residence, mother's education, mother's education, mother's occupation, vitamin A preparation supplements and diarrhoea.
The prevalence of SVAD in our study population was 6.9%, which was lower than the national average of 11.7%, but comparable with the 5.8% in the coastal areas in China in 2001Reference Lin, Liu, Ma, Tan, Zhang and Jiang13. Two other studies from Zhejiang and Fujian provinces which are adjacent to Anhui reported that the prevalence of SVAD was 7.8% in children aged 0–5 years and 6.3% in children aged 0–6 years, respectivelyReference Qiu, Chen, Yang, Huang, Ou and Chen15, Reference Yang, Chen, Chen, Zheng, Qiu and Chen16.
We found that breast-feeding in children, especially for those aged 1 year, was associated with an increased risk of SVAD. The effects remained statistically significant after adjustment for sex, age, the order in siblings, mother's education, mother's occupation, vitamin A supplementation and diarrhoea, as well as dietary factors. Univariate logistic regression analysis showed that those who received high, frequent intakes of complementary foods tended to be less likely to have SVAD. However, no food showed a statistically significant effect after adjustment for breast-feeding and other confounders. Residing in rural areas, ranking second or more in the order of siblings, mother's education less than junior high school, mother's occupation of farmer, no vitamin A preparation supplementation in the previous 6 months, and diarrhoea in the past 2 weeks were risk factors for SVAD in the univariate analysis; only children ranked second or more in the order of siblings were found to be at risk of SVAD in the fully adjusted logistic regression model.
Our finding regarding the association between breast-feeding and SVAD among children aged 0–5 years was comparable with a recent study of pre-school children in Nigeria, which shows that serum retinol deficiency was highest in those who were breast-feedingReference Oso, Abiodun, Omotade and Oyewole7. Another recent study in Kenya indicated that the risk of VAD in breast-fed infants older than 6 months was highReference Ettyang, Oloo, van Marken Lichtenbelt and Saris6. Generally, for a healthy and well-nourished mother, even exclusive breast-feeding for her baby during the first 6 months of life would be enough to provide all the nutrients including vitamin A17. Breast milk also protects babies from diarrhoea and acute respiratory infections that are considered to be the risk factors of VADReference Sommer and West3, Reference Alvarez, Salazar-Lindo, Kohatsu, Mirand and Stephensen18, Reference Salazar-Lindo, Salazar and Alvarez19. However, for malnourished mothers, breast-feeding alone may not meet the child's needs for vitamin A. Data reported over the past decade indicate that in developing countries, VAD may be even more common among women than in children, and deficient mothers may not provide sufficient vitamin A in their breast milkReference Katz, Khatry, West, Humphrey, Leclerg and Kimbrough8, Reference West9. Low breast milk concentrations of vitamin A were reported in lactating mothers in several studiesReference Katz, Khatry, West, Humphrey, Leclerg and Kimbrough8–Reference Gross, Hansel, Schultink, Shrimpton, Matulessi and Gross10. Another consideration is that if mothers failed to supply an adequate supplementary diet, the risk of SVAD for the breast-fed child would increase. Fawzi et al. reported that breast-fed children in the first 2 years of life were more likely to have a low dietary vitamin A intake compared with non-breast-fed children in the SudanReference Fawzi, Herrera, Nestel, el Amin and Mohammed11. High fertility with prolonged breast-feeding would further heighten the risk of VAD for both mothers and their children. Concentrations of vitamin A may not increase in accordance with the growth of children or with increased duration of lactationReference Gross, Hansel, Schultink, Shrimpton, Matulessi and Gross10. Our study also found that children ranked second or above among siblings were more likely to have SVAD. Therefore, latent factors underlying breast-feeding practice or breast milk should be considered as an explanation of the association between breast-feeding and SVAD in this study.
Because of increasing demands for vitamin A for rapid growth, breast milk alone may be not able to meet the entire needs of children aged 6 months or more. Other dietary sources of vitamin A should be added in their dietsReference Ettyang, Oloo, van Marken Lichtenbelt and Saris6. Vitamin A-rich foods from both animal and plant sources are recommended to be included in the diet for prevention of VADReference Sommer and West3. Our study showed that higher frequency intakes of milk, eggs, red and orange fruits, pulses and related products, dark green leafy vegetables, fish and shrimps were associated with reduced risks of SVAD, but not of meat and liver before adjustment for confounders. After adjustment for breast-feeding and other factors, no significant effect on SVAD was shown for those foods. In addition, the difference in risk of SVAD between breast-fed children and non-breast-fed children could not be fully explained by other dietary intakes. This may be partially due to the identical low consumption of expensive animal foods such as meat, liver and fish in the study communities. For plant sources of foods, such as fruit and vegetables, the bioavailability of pro-vitamin A carotenoids was poor (the bioavailability of β-carotene from vegetables and carrots was, on average, only a third of that of β-carotene in oil)Reference de Pee, West, Muhilal Karayadi and Hautvast20; therefore, the increased amount or frequencies of those foods may not ensure an obvious improvement in vitamin A status. Finally, the small numbers of SVAD cases in the study offered limited power for the test, while controlling confounders and covariates and the semi-quantitative food-frequency questionnaire used for this study only provided information about serving times of food categories but not exact quantities and specific names of foods.
The limitation of this study was that information on the intensity of breast-feeding which would provide further stronger evidence for the casual effects for breast-feeding was not collected. Nevertheless, this study provided results showing that breast-feeding, independently of other perceived factors, was associated with increased risk of SVAD in the children, especially in those with prolonged breast-feeding.
In conclusion, our study found breast-feeding in children, especially for those aged 1 year, could not ensure their protection from SVAD. This finding will help us to pay more attention to children with prolonged breast-feeding in the public health campaigns against VAD even in the areas where there is mild VAD, and consider more comprehensive measures to improve the vitamin A status in children besides breast-feeding. Furthermore, in order to understand the underlying reasons for the negative association between breast-feeding and SVAD, further research is needed to determine vitamin A status in lactating mothers, the retinol levels of their breast milk and their relationhips to SVAD in this province.
We are grateful to the children and their parents or caregivers for their participation in the study. We thank our field staff from the Institutes of Maternal and Children health care from Feixi, Shexian, Fentai, Guoyang county and Anqin City. We thank experts from Capital Institute of Pediatrics for their assistances in field technical guidance and analysing the serum retinol level of the children.