点击显示 收起
1 From the Center for Human Nutrition, The Johns Hopkins University School of Hygiene and Public Health, Baltimore, and the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B): Centre for Health and Population Research, Dhaka, Bangladesh.
2 Supported by cooperative agreements between The Johns Hopkins University School of Hygiene and Public Health, Baltimore, and the Office of Health and Nutrition, US Agency for International Development, Washington, DC (DAN-5116-1-00-8051-00 and HRN-A-00-97-00015-00). The study was a collaborative project between The Johns Hopkins University and the ICDDR,B: Centre for Health and Population Research. The ICDDR,B Centre for Health and Population Research is supported by agencies that share its concern for the health and population problems of developing countries. These are listed at the end of the text.
3 Address reprint requests to AL Rice, Division of Human Nutrition, Room W2041, The Johns Hopkins University School of Hygiene and Public Health, 615 North Wolfe Street, Baltimore, MD 21205. E-mail: arice{at}jhsph.edu.
See corresponding editorial on page 672.
ABSTRACT
Background: Conflicting results have been reported regarding the relative performance of serum retinol, the modified-relative-dose-response (MRDR) ratio, and breast-milk vitamin A concentrations in detecting changes in maternal vitamin A status.
Objective: We used receiver operating characteristic analyses and standardized differences to compare the ability of these indicators to detect a response to postpartum vitamin A supplementation in lactating Bangladeshi women.
Design: At 2 wk postpartum, women were randomly assigned to receive either a single dose of vitamin A [200000 IU (60000 retinol equivalents); n = 74] or placebo (n = 73). Data from maternal serum and breast milk collected 3 mo postpartum and from infant serum collected 6 mo postpartum were used to examine the ability of serum retinol, the MRDR ratio, and breast-milk vitamin A to discriminate between individuals in the supplemented and unsupplemented groups. Breast milk was collected by expressing the entire contents of one breast that had not been used to feed an infant for 2 h (full samples) or without controlling the time since the last breast-feeding episode (casual samples).
Results: Casual breast-milk samples performed better than full breast-milk samples in detecting a response to maternal supplementation. The MRDR ratio performed better than serum retinol in both the women and their infants. Overall, the most responsive indicator was the measurement of breast-milk vitamin A per gram of fat in casual breast-milk samples.
Conclusions: Breast-milk vitamin A and the MRDR ratio are responsive indicators of vitamin A status, especially in women with mild vitamin A deficiency.
Key Words: Vitamin A • supplementation • serum retinol • modified-relative-dose-response test • breast milk • lactation • women • Bangladesh
INTRODUCTION
Vitamin A deficiency is increasingly being recognized as an important public health problem among women in developing countries. Maternal supplementation, food fortification, nutrition education, and home gardening projects are some of the strategies being used to address this problem. However, standardized methods for evaluating the effectiveness of these interventions have not yet been developed, in part because appropriate indicators for assessing the vitamin A status of women for this purpose are not well defined.
The following indicators were used in previous studies to assess the vitamin A status of lactating women: serum retinol concentration, the relative-dose-response ratio, the modified-relative-dose-response (MRDR) ratio, conjunctival impression cytology, night blindness, and breast-milk vitamin A concentration, expressed either per volume or per gram of milk fat (1–10). These indicators have been used to characterize the extent of deficiency in populations by estimating the proportion of individuals with indicator values classified as abnormal or to detect differences in the vitamin A status of populations after an intervention. Indicators used for the latter purpose are called indicators of response (11).
Conflicting results have emerged regarding the relative performance of serum retinol, the MRDR ratio, and breast-milk vitamin A as indicators of response in women. The MRDR ratio was found to be more responsive than serum retinol in one study (9) but not in another (2). When comparing the responsiveness of the breast-milk indicators derived from samples collected 1 h after the last breast-feeding episode, one group of investigators found breast-milk vitamin A concentration per gram of fat to perform better than the concentration per volume (7), whereas other investigators found the opposite (2).
Because the vitamin A in breast milk is found almost exclusively in the milk fat globules, factors that affect breast-milk fat content affect the vitamin A concentration as well. A recent report by the World Health Organization (12) proposed 2 methods to either control or adjust for sampling-induced variations in breast-milk fat and vitamin A concentrations. First, to obtain representative breast-milk samples from individuals, the entire contents of one breast that has not been used to feed an infant for a set amount of time can be collected (full samples). For population-based studies, small amounts of breast milk can be collected from women at various times of day without controlling for the time since the last breast-feeding episode (casual samples). Alternatively, the fat content of breast-milk samples can be measured and the breast-milk vitamin A content expressed per gram of milk fat, rather than per volume. The effectiveness of these different methods for controlling sampling-induced variation in breast-milk fat and vitamin A concentrations has not been widely evaluated.
In the present report we used data collected in a postpartum maternal vitamin A supplementation trial conducted in a population of mildly deficient Bangladeshi women to compare the performance of serum retinol, the MRDR ratio, breast-milk vitamin A content per volume, and breast-milk vitamin A content per gram of milk fat as indicators of response. To examine the effect of the breast-milk sampling method on indicator performance, we also compared the breast-milk indicator data obtained from milk samples collected by the full and casual sampling methods.
SUBJECTS AND METHODS
Data collection
The data presented in this paper were collected as part of a randomized trial of maternal vitamin A and ß-carotene supplementation conducted in lactating women in Matlab, Bangladesh. The trial is described in detail elsewhere (13), but the design and results are summarized briefly here. At 2 wk (±1 wk) postpartum, women were randomly assigned to begin either vitamin A (one dose of 200000 IU [60000 retinol equivalents (REs)], followed by daily placebo capsules; n = 74), ß-carotene (daily capsules of 7.8 mg; n = 73), or placebo (daily capsules; n = 73) supplementation, which continued until 9 mo postpartum. Follow-up visits occurred at 3, 6, and 9 mo postpartum according to the schedule outlined in Table 1. Ninety-two percent of the participants completed the study. Both interventions had a positive effect on vitamin A status; the largest effect on maternal indicators was observed in the vitamin A group at 3 mo postpartum. For this reason, we chose to restrict the comparative analysis of the maternal indicators to the baseline and 3-mo postpartum time points. For simplicity, only data from the vitamin A and placebo groups are presented in this paper. The supplementation study was approved by The Johns Hopkins University School of Hygiene and Public Health Committee on Human Research and the Ethical Review Committee at the ICDDR,B: Centre for Health and Population Research.
View this table:
TABLE 1. . Data collection schedule for women participating in the supplementation trial
Breast-milk vitamin A and fat contents
During the supplementation study, breast-milk samples were collected from all women at 2 wk and 3, 6, and 9 mo (±1 wk) postpartum by either the full or casual collection method. For full collection, women came to the Matlab clinic and a staff member used a manual pump (White River, Laguna, CA) to help the women collect the entire contents of one breast that had not been used to feed an infant for 2 h. Casual breast-milk collection occurred during home visits when mothers manually expressed 5 mL breast milk into glass collection jars without controlling for the time since the last breast-feeding episode. Over the course of the study, each woman provided 2 full and 2 casual breast-milk samples according to a schedule that resulted in an equal distribution of the women in each treatment group to the 2 collection methods over time (Table 1).
Breast-milk vitamin A concentrations were assayed by the HPLC method described in detail elsewhere (13). The fat content of each breast-milk sample was measured in triplicate by using the creamatocrit method and Lucas et al's (14) equation for converting the volume measurement to grams of milk fat per liter milk. Breast-milk vitamin A content was calculated as the concentration per volume (µmol/L) and as the concentration per gram of milk fat (µg/g). The latter value was obtained by dividing the vitamin A concentration per volume (µg/L) by the fat concentration (g/L). According to criteria established by the World Health Organization, values 1.05 µmol/L and 8 µg/g fat were considered indicative of low breast-milk vitamin A content (12). Breast-milk data from the 2-wk and 3-mo visits were used for the analyses in this report.
Serum retinol concentrations and the modified-relative-dose-response test
During the supplementation study, the MRDR test was conducted at each visit on the 50% of women assigned to full breast-milk collection (Table 1) and on all infants at 6 mo of age. The field protocol for the MRDR test and the details of the HPLC assay used in measuring serum retinol concentrations and MRDR ratios are described elsewhere (13). MRDR ratios 0.06 were considered indicative of inadequate liver stores in both women and infants (12). Serum retinol concentrations <0.70 µmol/L and <1.05 µmol/L were considered indicative of vitamin A deficiency in infants and women, respectively (12, 15). Data from serum samples collected from women at 2 wk and 3 mo and from infants at 6 mo were used for the analyses in this report.
Statistical methods
The comparability of baseline characteristics in the vitamin A and placebo groups was examined by using Student's t test for continuous variables and the chi-square test for categorical variables. P values < 0.05 were considered statistically significant. To obtain normally distributed data for the t test, the values for breast-milk vitamin A content per volume, breast-milk vitamin A content per gram of fat, and maternal MRDR ratios were transformed by using the natural log function and the transformed values were tested. However, for ease of understanding, the untransformed values of the baseline data are reported. Statistical analyses were conducted with SPSS 7.5 for WINDOWS 95 (SPSS Inc, Chicago).
The performance of the indicators of response to supplementation was assessed by using standardized differences (11), calculations of sensitivity and specificity, receiver operating characteristic (ROC) curves (7, 16–18), and the corresponding areas under the ROC curves (19). Standardized differences were calculated from normally distributed data as d/SD, where d is the absolute difference in means between the 2 treatment groups and SD is the average of the SDs of the groups. The calculation for the standardized difference is related to the more familiar t test for between-group differences, which incorporates the observed sample size into the denominator of the equation (20). For the present analysis, the natural log–transformed values of breast-milk vitamin A content per volume, breast-milk vitamin A content per gram of fat, and maternal and infant MRDR ratios were used in the calculation of standardized differences. In this type of analysis, larger standardized differences represent more responsive indicators.
For the calculations of sensitivity and specificity, individuals in the placebo and vitamin A groups were considered as "diseased" and "nondiseased," respectively, and the standard formulas (21) were applied to nontransformed data. Sensitivity and specificity are traditionally defined as the probabilities that a test (an indicator with a given cutoff value) will correctly identify the diseased and nondiseased individuals, respectively. Sensitivity was calculated as TP/(TP + FN) and specificity was calculated as TN/(TN + FP), where TP is true positive [placebo group members with a positive test result (ie, low breast-milk vitamin A)], FN is false negative [placebo group members with a negative test result (ie, adequate breast-milk vitamin A)], FP is false positive (vitamin A group members with a positive test result), and TN is true negative (vitamin A group members with a negative test result).
ROC curves were plotted to compare graphically the ability of the indicators to discriminate between the vitamin A and placebo groups over a range of vitamin A status. The curves were constructed by using the sensitivity and specificity data obtained when the cutoff for each indicator was varied over the entire range of indicator results. The data were plotted as sensitivity versus 1 - specificity. In this type of plot, points in the lower left portion of the curve reflect cutoffs at lower vitamin A status and those in the upper right portion reflect cutoffs at higher status. When comparing indicators, the higher the sensitivity is for any given level of specificity, the more responsive the indicator is at that level of status.
The diagonal line running from the lower left to the upper right corner of the ROC curve is called the line of indifference and represents an indicator that does not discriminate between groups better than chance alone. In qualitative terms, the closer a curve is to the upper left corner of the graph, the better the indicator discriminates between the 2 groups. The curve for a perfectly discriminating indicator passes through the point of 100 - specificity and 100 - sensitivity. Indicators that are equally responsive over the entire range of cutoff values in a data set are represented by a symmetric curve.
The area under the curve and corresponding 95% CI was calculated for each indicator of response as a quantitative measure of discriminating ability (16). Indicators that discriminate perfectly between groups have an area of 1.0, whereas those that discriminate no better than chance alone have an area of 0.5. To compare the performance of indicators measured in the same individuals, the difference in the areas under the curve was also calculated for the following pairs of indicators: maternal MRDR ratio and maternal serum retinol concentration, infant MRDR ratio and infant serum retinol concentration, breast-milk vitamin A content per volume and breast-milk vitamin A content per gram of fat in casual breast-milk samples, and breast-milk vitamin A content per volume and breast-milk vitamin A content per gram of fat in full breast-milk samples. The MEDCALC software package (version 4.0; MedCalc, Brussels) was used to create the ROC curves and to calculate and test differences in the areas under the curves (22).
RESULTS
At baseline, the groups of women randomly assigned to the vitamin A and placebo treatments were not significantly different with respect to any of the characteristics shown in Table 2. The assessment of maternal vitamin A status suggested that this population of women was mildly vitamin A deficient at 2 wk postpartum. In the 2 groups combined, 13% of the women had serum retinol concentrations <1.05 µmol/L and 17% had MRDR ratios indicative of low liver stores. Although population criteria for these indicators have not been established for women, the observed prevalences of low serum retinol concentrations and positive MRDR ratios are consistent with a mild public health problem according to the World Health Organization criteria for preschool-age children (12).
View this table:
TABLE 2. . Baseline characteristics of women at 2 wk postpartum1
The baseline prevalences of low breast-milk vitamin A in the 2 groups combined were 44% and 37% for full breast-milk samples when the vitamin A concentration was expressed per volume and per gram of fat, respectively. For casual breast-milk samples, these prevalences were 24% and 46%, respectively. The criteria for low breast-milk vitamin A in mature breast milk were applied to the baseline samples, even though the women were secreting transitional rather than mature breast milk at 2 wk postpartum (23), because separate criteria for transitional breast milk do not exist. The observed prevalences are indicative of a severe public health problem (12). Because breast-milk vitamin A declines rapidly from high concentrations in colostrum to lower concentrations in mature breast milk (24), the magnitude of the underlying problem may be more severe than indicated by these transitional breast-milk values.
The response to postpartum maternal vitamin A supplementation is shown in Table 3, in which the indicators are listed in descending order of their ability to discriminate between the 2 treatment groups 2.5 mo after supplementation. On the basis of the absolute values of the standardized differences, the most responsive indicator was breast-milk vitamin A content per gram of fat in casual breast-milk samples. The maternal MRDR ratio, breast-milk vitamin A content per volume in casual samples, and the infant MRDR ratio were somewhat less discriminating, followed by serum retinol concentrations. The least discriminating indicators were the breast-milk vitamin A indicators from full breast-milk samples.
View this table:
TABLE 3. . Response to postpartum maternal vitamin A supplementation measured by maternal and infant indicators
The performance of the breast-milk indicators varied by the type of breast-milk sample. Breast-milk vitamin A content per gram of fat performed better than breast-milk vitamin A content per volume in casual but not full breast-milk samples. The performance of breast-milk vitamin A content per gram of fat could be biased if breast-milk fat concentrations were related to the treatment, but this was not the case. At 3 mo postpartum, the fat content of the breast milk was lower in the full samples than in the casual samples, but did not differ significantly by treatment group. In full breast-milk samples, the fat content was 34.0 ± 13.3 g/L (
The ability of the indicators to discriminate between the vitamin A and placebo groups over a range of vitamin A status is shown in a series of ROC graphs (Figures 1–4). In qualitative terms, curves further away from the line of indifference represent more responsive indicators at a given vitamin A status. In these plots, points in the lower left portion of the curve reflect lower vitamin A status whereas those in the upper right portion reflect higher status. To provide a frame of reference, selected points on the graph have been labeled with the indicator values.
FIGURE 1. . Receiver operating characteristic curves comparing the modified-relative-dose-response (MRDR) ratio and serum retinol concentrations in women at 3 mo postpartum as indicators of response to maternal vitamin A supplementation. For the MRDR ratio, the suggested cutoff of 0.06 is highlighted (12). The serum retinol cutoffs of 1.05 and 1.40 µmol/L are also highlighted for comparison purposes.
FIGURE 2. . Receiver operating characteristic curves comparing the modified-relative-dose-response (MRDR) ratio and serum retinol concentrations in infants at 6 mo postpartum as indicators of response to maternal vitamin A supplementation. The commonly used cutoffs for the MRDR ratio (0.06) and serum retinol (0.70 µmol/L) in this age group are highlighted (12). The other highlighted values are included for comparison purposes.
FIGURE 3. . Receiver operating characteristic curves comparing breast-milk vitamin A content per volume and per gram of fat in full breast-milk samples at 3 mo postpartum as indicators of response to maternal vitamin A supplementation. The World Health Organization suggested cutoffs for these indicators are highlighted (12).
FIGURE 4. . Receiver operating characteristic curves comparing breast-milk vitamin A content per volume and per gram of fat in casual breast-milk samples at 3 mo postpartum as indicators of response to maternal vitamin A supplementation. The World Health Organization suggested cutoffs for these indicators are highlighted (12).
When assessed in this way, the MRDR ratio performed better than serum retinol in both women (Figure 1) and their 6-mo-old infants (Figure 2). In women, the MRDR ratio was most responsive at low vitamin A status, whereas serum was responsive only at higher status. Most women in both the vitamin A and placebo groups had apparently adequate serum retinol concentrations. At 3 mo postpartum, the overall mean was 1.40 µmol/L and 80% had concentrations 1.05 µmol/L. In contrast, the mean serum retinol concentration of infants in both groups combined was 0.80 µmol/L and nearly all (90%) had MRDR ratios indicative of low liver stores. At this overall lower vitamin A status, the responsiveness of the MRDR ratio did not differ markedly across the range of observed values.
The ability of breast-milk vitamin A content per volume and per gram of fat to discriminate between the treatment groups is shown for full (Figure 3) and casual (Figure 4) breast-milk samples separately. In full breast-milk samples, neither indicator was very informative (ie, curves for both were close to the line of indifference). In casual breast-milk samples, both indicators were responsive and breast-milk vitamin A content per gram of fat was more responsive than breast-milk vitamin A content per volume at low status.
DISCUSSION
Conflicting results have emerged from previous studies of the performance of serum retinol, the MRDR ratio, and breast-milk vitamin A concentrations as indicators of response. These apparently conflicting findings may be partially explained by differences among the studies such as the initial vitamin A status of the population, the type of intervention, the duration of the study, the dosage used, the timing of specimen collection, the study design, the statistical treatment of the variables studied, and the reference point with which the indicators were compared (25, 26). In addition, coexisting nutritional deficiencies or infections may have affected the performance of these indicators differently, but such effects have not been well described.
Our analysis of the postintervention data from the trial in Bangladesh showed that the indicators we evaluated differed in their ability to detect a response to postpartum maternal vitamin A supplementation. This analysis relied on the randomized nature of the trial to ensure that there were no important baseline differences in the vitamin A status of the women who contributed data to the 3-mo follow-up visit. If such imbalances existed, our conclusions about the responsiveness of the indicators could be biased. However, the available data do not suggest that this occurred. There were no statistically significant between-group differences in the baseline characteristics assessed in all women nor were there significant differences in the vitamin A status indicators for the subset of the women who completed the same type of visit at the baseline and 3-mo time points (data not shown).
In this study, the lactating women were mildly vitamin A deficient at 2 wk postpartum and responded to supplementation with one 200000-IU (600000-RE) dose of vitamin A. We found that the MRDR ratio was more responsive than serum retinol, especially in women with low vitamin A status. The standardized differences and areas under the ROC curve showed that the MRDR ratio could detect differences in vitamin A status even when serum retinol concentrations fell in a range generally considered adequate for adults.
Our findings are consistent with the theoretical mechanism involved in the MRDR test. The MRDR ratio assesses liver vitamin A stores by measuring the buildup of apo-retinol binding protein (apo-RBP) that occurs as vitamin A stores decline (8). Although serum retinol concentrations are homeostatically controlled over a wide range of liver stores, the appearance of the vitamin A2 analog bound to accumulated apo-RBP, which occurs during the MRDR test, appears to be less tightly regulated. Therefore, the MRDR ratio should detect differences in vitamin A status when serum retinol concentrations do not.
Among infants, the MRDR ratio was also a more discriminating indicator than serum retinol. However, in contrast with that in the women, the responsiveness of the MRDR ratio in the infants did not differ markedly over the observed range of vitamin A status. Many infants had MRDR ratios that far exceeded the cutoff (0.06) considered indicative of low liver stores. In addition, the mean serum retinol concentration for infants was lower and the range of values was smaller than that observed in the women. The increase in breast-milk vitamin A intake may have been large enough to cause a response in the infant's MRDR ratio, but insufficient to build liver stores to a level that would influence serum retinol concentrations. Alternatively, the serum retinol concentrations of these infants may have been high enough to be under homeostatic control.
In a study in which Indonesian children aged 6 mo to 6.6 y with serum retinol concentrations of 0.80 µmol/L were supplemented with 210 µmol retinyl palmitate, a response toward improved MRDR ratios was also observed in the absence of a significant increase in serum retinol concentrations (27). The authors noted that the circulating serum retinol concentrations of Asian children are quite low in comparison with those of American children and suggested that secondary factors, such as macronutrient or micronutrient deficiencies or infections, may also act to maintain serum retinol concentrations at an overall lower level.
The finding that the responsiveness of breast-milk vitamin A content per volume and per gram of fat differed markedly according to the breast-milk sampling method was unexpected. We had hypothesized that the full sampling method would result in samples that were more representative of individual women, but that at a population level, both the full and casual sampling methods would produce groups of samples that were equally effective in detecting a response to the intervention. This was not the case. Although both sampling methods and both indicators showed that women who received vitamin A produced breast milk with a higher vitamin A content than that in their unsupplemented counterparts, the observed differences were much smaller in the full breast-milk samples.
This finding may be explained by the variability in the fat content and therefore in the vitamin A content of the breast-milk samples collected. Throughout the study, the more controlled protocol of full breast-milk collection produced samples consistently lower in fat than did the casual sampling protocol. The fat content of the full breast-milk samples was similar to results published for mature breast milk (33 g/L) (23), whereas the fat content of the casual breast-milk samples was 35% higher. When the breast-milk samples were stratified by fat content rather than collection method, the treatment groups remained balanced with respect to fat content but the standardized differences were larger for the higher-fat samples. In these samples (containing >36 g fat/L, the median value for all breast-milk samples), the standardized differences between the treatment groups were 0.81 and 0.78 for the natural log–transformed values of breast-milk vitamin A content per volume and per gram of fat, respectively. For the lower-fat samples (those containing 36 g fat/L), the corresponding standardized differences were 0.04 and 0.05.
Our analysis showed that the indicators studied differed in their ability to detect a response to postpartum maternal vitamin A supplementation. We conclude that the MRDR ratio is more efficient than serum retinol in detecting a response to an intervention in populations of women in whom serum retinol concentrations are close to the homeostatically controlled range. In infants, the MRDR ratio is more responsive than serum retinol among populations with extremely poor liver stores, but only marginally low serum retinol concentrations. However, when the dynamics of vitamin A storage and transfer are of interest, both of these indicators will provide valuable information.
Our evaluation of the breast-milk indicators supports the World Health Organization recommendation that the logistically simpler method of casual breast-milk collection can be used to obtain data for evaluating interventions targeted to women (12). We found that the indicators derived from casual breast-milk samples were more informative as indicators of response than were those derived from full breast-milk samples. In the casual samples, expressing breast-milk vitamin A content per gram of fat rather than per volume improved the discriminating ability of the indicator. Because breast-milk fat content is relatively easy to measure by the creamatocrit method (14), even under field conditions, it should be measured whenever possible.
ACKNOWLEDGMENTS
We thank J Chakraborty, the RETIBETA staff, and the study participants in Matlab, Bangladesh, for their enthusiasm and cooperation during the fieldwork phase of the supplementation trial; the staff of the Biochemistry and Nutrition Laboratory at the ICDDR,B: Centre for Health and Population Research for analyzing the breast-milk samples; and the laboratory staff at the Center for Human Nutrition at The Johns Hopkins University for analyzing the serum samples. Supporters of the ICDDR,B: Centre for Health and Population Research include the following: the aid agencies of the governments of Australia, Bangladesh, Belgium, Canada, the European Union, Japan, the Netherlands, Norway, Saudi Arabia, Sweden, Switzerland, the United Kingdom, and the United States; the United Nations Development Program; the United Nations Children's Fund; the World Health Organization; the International Atomic Energy Agency; the International Center for Research on Women; the International Development Research Center; the Population Council; the Swiss Red Cross; the World Bank; the Aga Khan Foundation; the Child Health Foundation; the Ford Foundation; the George Mason Foundation; the Rockefeller Foundation; the International Life Sciences Institute; the National Institutes of Health; the New England Medical Center; Northfield Laboratories; Procter and Gamble; Rhône Poulenc Rorer; the Thrasher Research Fund; Johns Hopkins University; Karolinska Institute; Loughborough University; London School of Hygiene & Tropical Medicine; University of Alabama at Birmingham; University of Goteborg; University of Pennsylvania; University of Virginia; American Express Bank; Helen Keller International; Lederle Praxis; NRECA International Ltd; The Rand Corporation; Save the Children Fund-USA; Social Development Center of the Philippines; UCB Osmotics Ltd; and Wander AG.
REFERENCES