Vitamin A deficiency (VAD) is a major micronutrient problem in many developing countries. There are estimates that worldwide approximately 140 million children of pre-school age suffer from VAD every year and of these 4.4 million suffer from xerophthalmia1, Reference Sommer and Davidson2. Nearly half of all VAD occurs in South and Southeast AsiaReference Underwood, Sporn, Roberts and Goodman3. VAD might increase children's susceptibility to infection, reduce physical growth and decrease the possibility of survival from serious illness. A number of epidemiological studies have identified an association between VAD and higher rates of mortalityReference Schultink4, Reference Fawzi, Herrera and Spiegelman5.
VAD in African countries is associated with insufficiently varied diets, poor maternal education and low economic statusReference Schemann, Banou, Guindo, Joret, Traore and Malvy6, Reference Kassaye, Receveur, Johns and Becklake7. Young children are susceptible to VAD if their dietary intake of the essential vitamin is inadequate. Although the prevalence of protein–energy deficiency malnutrition has decreased in most Chinese areas with economic development in recent yearsReference Chang and Xian8, Reference Chen, He, Hu, Fu and Wang9, micronutrient deficiency might still be an issue. A study from western China showed that VAD was a health problem among children that had affected their growthReference Hu, Tong, Oldenburg and Feng10. At the 1990 World Summit for Children, the World Health Organization (WHO), the United Nations Children's Fund and governments of the world committed themselves to eliminate VAD as a public health problem by the year 2000. Elimination of VAD was declared a National Programme of Action for Child Health in China in the 1990s. In order to obtain baseline information on vitamin A status and identify an intervention strategy, the first cross-sectional survey of VAD was conducted from November 1999 to July 2000 in the whole of China. The present paper estimates vitamin A status and the prevalence of VAD, and analyses its relationship to sociodemographic and dietary factors.
Methods
Study population
Data from 8681 children aged 0–6 years were collected. The sampling design consisted of a three-step stratified random sampling. In the first step, four or five provinces were selected at random in each of the three geographic areas defined by the State Statistics Bureau of China11 – coastal, inland and west. In the second step, within each province one city and two counties were selected at random. In each city, two districts and within the districts three communities were selected at random. In rural areas three townships and within townships two villages were selected at random (Fig. 1). In each community or village, 11 or 12 children in each age group (0–23 months, 24–47 months and 48–71 months) were randomly selected. Overall, less than 2% of the sample was classified as a non-responder due to parental refusal to participate in the study or failure to provide a blood sample for analysis.
Sampling design
Since information on vitamin supplements was provided by only 8% of parents, these children were excluded from the analysis. Additionally we excluded children having a disease in the two weeks immediately before the examination. Thus the analysis was confined to 7826 healthy children with complete information and plausible information on explanatory variables and vitamin A level. Furthermore, in the first year of life, a large amount of heterogeneity is likely due to variable initiation and duration of breast-feeding and differences in the onset of introducing solids. We therefore confined the multivariate analyses to a sub-sample of 6885 children aged 1–6 years. The proportion of children from cities versus children from rural areas was 1:3 (Fig. 1). This survey was approved by the Ministry of Health of China and parental consent was obtained for all children before enrolment.
Measurements
A 200 μl capillary blood sample was taken from each child by puncture of the left ring finger. The blood samples were protected from light, stored in cool boxes and centrifuged for 10 min within 4–6 h after collection. The serum was protected from light and stored at − 20°C for up to 9 days and then transported for a maximum of 6 h in a cool box before it was stored at − 70°C in the laboratory of the Capital Institute of Pediatrics in Beijing, where the serum samples were analysed within 3 months.
The serum samples were analysed for retinol using fluorescence microanalysis under a strict quality control protocol. A consultant was sent from the Centers for Disease Control and Prevention (CDC), Atlanta, GA, USA to direct the serum retinol measurements at the Capital Institute of Pediatrics, Beijing. The fluorescence spectrophotometer was an RF-540 from Shimadzu (Kyoto, Japan). Standard serum samples (n = 100) from CDC, Atlanta were analysed for serum retinol as a quality control measure. The concentration of retinol in the standard serum samples, as measured by high-performance liquid chromatography assay (HPLC), was 1.336 μmol l− 1. The acceptable range of measurement for the serum retinol concentration was 1.225–1.575 μmol l− 1. The range of serum retinol concentration for the 100 serum samples measured by fluorescence microanalysis was 1.278–1.545 μmol l− 1. The coefficient of variation for repeat measurements was 2.9–9.9%.
Serum retinol concentration of < 0.7 μmol l− 1 was used as a cut-off point for VAD and serum retinol concentration < 0.35 μmol l− 1 as the cut-off point for severe VAD, in accordance with widely used international definitionsReference Sommer and Davidson2, Reference West12 and the definition recommended by the Chinese Association of Pediatrics13. Serum retinol concentration < 1.05 μmol l− 1 was defined as marginal VADReference Underwood, Sporn, Roberts and Goodman3.
Interview
The interview on diet was based on the Helen Keller International Food Frequency Method to obtain information on the frequency of consumption of dietary sources of vitamin A intake of the children. The mothers or caregivers were asked how often their children had milk, liver, meats (all kinds of meats excluding liver), fish, eggs and dark green (orange) fresh vegetables in the past week. The Helen Keller Food Frequency Method has been validated to provide an accurate community-level measure for identification of a diet associated with VAD and has been used elsewhereReference Sloan, Rosen, de la Paz, Arita, Temalilwa and Solomons14, Reference Hu, Rimm, Smith-Warner, Feskanich, Stampfer and Ascherio15. Additionally, information on parental education and ethnicity was obtained within the interview. Education was categorised into four levels: primary school not completed (incomplete or no primary school), primary school (graduated from primary school), high school (graduated from high school or incomplete high school) and college or above (graduated from college or above or incomplete college). Ethnicity was categorised as Han (majority) and non-Han (minority ethnicity). The questions were pre-tested; after the initial training course, each of 28 participating physicians interviewed 10 individuals for pre-testing. The questions were then modified to increase practicability, reliability and validity. The interviews were performed by two trained local doctors and one person from the study team at each field site. In all households, the primary respondent was the mother of the selected child or the principal caregiver.
Statistical methods
The prevalence of VAD was estimated with 95% exact confidence limits based on the binomial distributionReference Clopper and Pearson16. A multiple linear regression model was used to assess the impact of potential predictors on serum retinol concentrations (μmol l− 1). Additionally a logistic regression model was estimated with VAD defined at a serum retinol concentration of < 0.7 μmol l− 1. All covariates associated with serum retinol concentrations or VAD were considered for a multiple logistic or a linear regression model, respectively. Only significant (P < 0.05) predictors were included in the final models.
Outliers (data points which do not appear to follow the characteristic distribution of the rest of the data) were excluded from the analysis to minimise the impact of measurement errors or other anomalies. For each individual outcome measurement we tested whether its exclusion changed the standard deviation of the outcome variable by more than 10%. Any observation fulfilling this criterion was excluded as an outlier. Regression diagnostics further included examination of residuals, non-constant error of variance (heteroscedasticity) and non-linearity. Multicollinearity of respective covariates was identifiedReference Belsley, Welsch and Kuh17 by a variance inflation factor >10.
All calculations were carried out with the software package SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).
Results
VAD (serum retinol < 0.7 μmol l− 1) was observed in 957 out of 7826 children (12.2% of the entire study population), whereas severe VAD (serum retinol < 0.35 μmol l− 1) was observed in 39 (0.5%). Serum retinol concentrations were marginal (0.7 to < 1.05 μmol l− 1) in 3140 (40.1%) children aged 0–6 years old.
The highest prevalence of VAD (serum retinol < 0.7 μmol l− 1) was observed in children of mothers without Han ethnicity (22.7%) or lower maternal education (primary school not completed) (19.8%) and in children living in the poor western area (17.4%) (Table 1). The prevalence of VAD showed an inverse relationship to children's age, level of maternal education and region, with an increase in the order of coastal area followed by the inland area and the western area (Table 1).
Prevalence of VAD (serum retinol <0.7 μmol l−1) among Chinese children
* Cochran–Armitage trend test: P < 0.0001.
The left-shifted lower percentiles for serum retinol concentration in the various strata were not due to skewed distributions of the serum retinol levels. Rather the explanatory variables were related to shifts of the entire distribution of the serum retinol levels. As an example, the association between Han ethnicity and serum retinol concentration is shown in Fig. 2. In children of Han mothers, the crude mean serum retinol concentration was 0.16 μmol l− 1 higher than in children of non-Han mothers.
Effect of mothers' ethnicity on serum retinol concentration among Chinese children aged 1–6 years: (a) Han ethnicity (n = 6062; serum retinol concentration 1.09 ± 0.31 μmol l− 1, mean ± standard deviation, SD); (b) non-Han ethnicity (n = 1764; serum retinol concentration 0.93 ± 0.31 μmol l− 1, mean ± SD). Unadjusted mean difference: − 0.16 μmol l− 1
In order to assess the mutually independent relationships between the explanatory variables depicted in Table 1 and nutritional intake, the mutually adjusted increment in serum retinol related to each of these explanatory variables is shown in Table 2 for children aged 1 year and older. The child's age, coastal geographic area, high maternal education and Han ethnicity were independently associated with higher vitamin A levels for children.
Significant increases (P<0.05) in mean retinol concentration (multiple linear regression) and significant (P<0.05) adjusted OR for VAD (serum retinol <0.7 μmol l−1; multiple logistic regression) among n=6885 children aged 1–6 years
OR – odds ratio; VAD – vitamin A deficiency; SD – standard deviation; CI – confidence interval.
* Adjusted for all variables displayed in table.
† Adjusted for geographical area, maternal ethnicity, maternal school education and consumption of milk, eggs and meat.
Similar results were observed for VAD as categorical outcome variable (multiple logistic regression analysis): independent protective effects of Han nationality, high maternal education, coastal area and additional year of life were observed which could not be explained by the consumption of milk, meat or eggs (Table 2). The adjusted odds ratio of belonging to a minority, having a mother with poor education and living in the poor western area, compared with belonging to Han nationality, having a highly educated mother and living in coastal parts of the country (reference), for example, was 11.6 ( 11.6 = 1 / exp [ ln ( 0.54 ) + ln ( 0.30 ) + ln ( 0.53 ) ] ; see Table 2). This was due to a lowered average serum retinol level of − 0.29 μmol l− 1 ( − 0.29 = − 0.08 + ( − 0.16 ) + ( − 0.05 ); see Table 2).
Discussion
The main finding of this study was an average prevalence of VAD of 12.2%, including 0.5% of severe VAD in Chinese children aged 0–6 years. While the VAD prevalence in the eastern coastal areas was only 6.2%, it was almost three times as common in the western parts of China. These figures are close to those found in a previous study with small sample sizeReference Qin, Ge and Dai18. In the present survey, belonging to an ethnic minority, having a mother with poor education and living in the poor western area accounted for an adjusted odds ratio of 11.6 for VAD compared with belonging to HAN ethnicity, having a highly educated mother and living in coastal parts of the country.
The strengths of this study are a defined random sampling procedure and the use of a standardised method to measure VAD. The study sites covered half the provinces in China. They are likely to reflect the prevalence of VAD in the whole of China because the distribution of socio-economic classes for the selected provinces was similar to that of the national distribution according to the Classification of Chinese Health and Economic Status 19.
Vitamin A status can be assessed by ophthalmologic examination, measuring serum retinol concentration and the Modified Relative Dose Response. Serum retinol concentration is usually measured by HPLC. For the sake of practicality in the present large epidemiological study, fluorescence microanalysis was used to measure the serum retinol level. The pre-testing showed similar reliability and validity for fluorescence microanalysis as for HPLC. The cut-off values for VAD (serum retinol < 0.35 μmol l− 1, < 0.7 μmol l− 1 and 0.7 to < 1.05 μmol l− 1) were chosen as recommended by WHO1. Severe VAD can be defined as serum retinol < 0.35 μmol l− 1 or ophthalmologic symptomsReference Sommer and Davidson2. Serum retinol < 1.05 μmol l− 1 was considered indicative of potentially suboptimal vitamin A statusReference Sommer and Davidson2, 13. It appeared that the prevalences of VAD and potentially suboptimal concentrations of serum retinol were high in children in China, whereas the prevalence of severe or clinically manifest VAD was very low.
VAD is a common nutritional disorder in many developing countries; e.g. in northern Ethiopia, VAD prevalence is 59.5% in children aged 6–9 yearsReference Kassaye, Receveur, Johns and Becklake7 and in Nepal the prevalence is 32.3% among pre-school childrenReference Gorstein, Shreshtra, Pandey, Adhikari and Pradhan20. Compared with these and other developing countriesReference Schemann, Banou, Guindo, Joret, Traore and Malvy6, Reference Ngare, Muttunga and Njonge21, the average prevalence of VAD was lower in China. China is a big country with unbalanced economic development. Although VAD prevalence was only 6.2% in the rich coastal area, which is comparable to data from the USAReference Ballew, Bowman, Sowell and Gillespie22, VAD still remains a public health problem in the poor western area where its prevalence was 17.4%.
The available dietary data could not fully explain different serum retinol levels between the various socio-economic subgroups. The consumption of all animal foods (liver, milk, meat, fish and eggs) is associated with serum retinol concentration. Usually, animal foods (including vitamin-A-rich animal foods) are more expensive than vegetables in China. The higher consumption of animal foods may be a proxy for a higher income or a better diet. It might reflect the effect of dietary intake on vitamin A level indirectly. Dark green vegetables are a major source of vitamin A but, surprisingly, the consumption of dark green vegetables in the week prior to the examination had no relationship with serum retinol concentration. This might be due to a lack of adequate variation in consumption among this population – 82% had eaten such vegetables daily. The diet in China may be too different from that of those countries where the Helen Keller questionnaire was validatedReference Sloan, Rosen, de la Paz, Arita, Temalilwa and Solomons14.
A significant ethnic difference in serum retinol concentration was observed in this study. The prevalence of VAD was 22.7% in children of non-Han mothers compared with 9.2% in children of Han mothers. According to WHO criteria, a prevalence of 20% or above for serum retinol < 0.7 μmol l− 1 is indicative of a VAD problem of public health significanceReference Underwood and Arthur23. These WHO criteria were based on surveys of children aged 6–71 months. The age range within this survey was similar.
Most of minority groups live in the western parts of China. Their economic status is below that of the Han majority: the per capita income of non-Han people is 80% of the per capita income of Han people24. Surprisingly we found a strong and significant impact of region on the risk for VAD, even after adjustment for maternal education and minority status. There are plenty of data on the socio-economic differences between eastern and western areas: per capita income in the western area is only 71% of the per capita income in eastern urban areasReference Stephensen and Gildengorin25. Unfortunately, we lacked individual data on income in relation to regional living costs. This might account for a bias towards unity due to random misclassification of the exposure.
Younger children had a higher prevalence of VAD than older children. This finding is consistent with other reportsReference Stephensen and Gildengorin25. It might also be due to the use of one single cut-off point for all age groups, not taking into account potentially physiologically lower vitamin A levels in younger children. Unfortunately, no international age-specific cut-off points for low serum retinol concentration exist for young children. The inverse relationship between age and low serum retinol concentration might be attributed to long-term effects of vitamin-A-rich foods or supplements. Breast milk is a better source of vitamin A for infants than formula feeding. Breast-feeding and fortified complementary feeding should be improved constantly for infants and young children in order to increase the accumulation of nutrients, including vitamin A.
Conclusion
A high prevalence of VAD was observed in western provinces compared with eastern areas in China. Particularly minority children and those with mothers having poor education have been identified as risk groups of public health importance. These children should be targeted for preventive measures.
A strategy of vitamin A supplementation has been put into the National Programme of Action for Child Health in China (2001–2010) following the findings of the present national survey. Additionally, after this study vitamin A and iron folic acid supplementation to pregnant women was implemented in some rural provinces in China. However, there is no official health-care policy generally targeting new mothers at children's vitamin A supplementation. Thus, a comprehensive, long-term national strategy is needed to promote sustained improvement in vitamin A status in China. Vitamin A supplementation should be offered during public vaccination consultation hours, regular consumption of vitamin-A-rich foods should be promoted by nutrition education, and breast-feeding and fortified complementary feeding in infants and young children should be improved.
Acknowledgements
This survey was carried out with the aid of a nutrition grant (YH101-13) from UNICEF. We thank the mothers and children who participated in this study, Ma Guanfu, Tan Zangwen and Liu Min in the Capital Institute of Pediatrics for their laboratory analyses of serum retinol, Liu Yulin for her advice and the members of our field team. Additionally we want to thank the editorial team and the reviewers for their helpful comments to improve the manuscript.
