The nutrition transition, underlined by economic development, modernization and rapid urbanization( Reference Popkin 1 ), has perpetuated the disproportionately high burden of non-communicable diseases (NCD) in low- and middle-income countries( Reference Popkin 2 ). Patterns of diet and physical activity have shifted towards increased consumption of high-energy, nutrient-poor foods and increased sedentary activity, resulting in chronic energy imbalance and poor nutritional outcomes( Reference Popkin 3 ). The adverse health effects associated with this nutrition transition are often greatest in children under 5 years of age( Reference Drewnowski and Popkin 4 ). Changes in lifestyle often threaten appropriate development and growth in the earliest life stages, contributing to a higher risk of malnutrition in all its forms (under- and overnutrition)( 5 ) and premature NCD mortality( Reference Anderson 6 ). Particularly in young children, malnutrition has been closely linked with increased mortality( Reference Pelletier, Frongillo and Schroeder 7 ) and impaired cognitive, physical and metabolic development( Reference Walker, Wachs and Gardner 8 ).
Samoa is a low- and middle-income country in the South Pacific undergoing significant nutrition transition, with shifts in societal and dietary structures along with significant rises in NCD prevalence over the past two decades( Reference Baker, Hanna and Baker 9 – Reference McGarvey 11 ). High levels of obesity and other associated metabolic disorders have been well documented among the Samoan adult population( Reference Hawley, Minster and Weeks 12 – Reference DiBello, McGarvey and Kraft 14 ). A high prevalence of overweight and obesity has also been observed among school-aged children. Data from 2003 showed 14·8 % of boys and 6·3 % of girls aged 6–8 years to be overweight, while 3·3 % of boys and 14·3 % of girls were already obese( Reference Keighley, McGarvey and Quested 10 ). While the burden of NCD is clear among adults and school-aged children, there remains a paucity of data about nutritional status in children under 5 years of age.
Present knowledge of child nutritional status in this age group relies on data from the Samoa National Nutrition Survey conducted over 15 years ago (1999)( Reference Mackerras and Kiernan 15 ). Data from that survey showed that 4·2 % of children aged 6–59 months were stunted and 1·9 % were underweight. At the same time, 10·6 % of children aged 24–59 months were overweight or obese and 23·2 % were anaemic( Reference Mackerras and Kiernan 15 ). Since dramatic rises in obesity and concurrent problems with malnutrition have been identified in Samoan adults and school-aged children over the past 20 years( Reference Keighley, McGarvey and Quested 10 , Reference Hawley, Minster and Weeks 12 , Reference Seiden, Hawley and Schulz 13 , 16 ), a reassessment was warranted in young children.
The present study served to document the current nutritional status of Samoan children between 24 and 59 months old and to identify child, maternal and household-level correlates of malnutrition. The specific study aims were to: (i) to describe the prevalence and coexistence of four indicators of malnutrition, namely stunting, overweight/obesity, anaemia and underweight; (ii) compare the prevalence of these nutritional indicators in three census regions with varying levels of urbanization and exposure to the nutrition transition; and (iii) examine associations between selected child, maternal and household characteristics and child nutritional status. Understanding the correlates of malnutrition is critical to informing future NCD control and prevention strategies targeted at this age group.
Methods
Study population
The recruitment and measurement of study participants took place from June to August 2015 on the island of Upolu, Samoa. The majority (76 %) of the Samoan population resides on Upolu, with the rest of the population residing on the more rural island of Savai’i and other smaller, outlying islands. From 285 villages within Upolu, ten villages were included in the study: four from the Apia Urban Area (AUA; an urban region), three from Northwest Upolu (NWU; a peri-urban region) and three from Rest of Upolu (ROU; a rural region). These villages were selected based on their prior participation in an adult obesity study( Reference Hawley, Minster and Weeks 12 ). Exposure to the nutrition transition closely follows the gradient of urbanization across these regions, with AUA the most exposed and ROU the least. For the ease of data collection, we focused on large villages (population >500), paying close attention to geographic distribution across each region. The Ministry of Health and the Samoan Bureau of Statistics arranged for village-level permission to spend 2 to 3 d in each village and complete study activities. Recruitment was undertaken with the cooperation of village leaders and women’s committee representatives, whose role was to explain the purpose and procedures of the study to women residing in the village and promote interest in participation. In each village, any mother and child pair that wanted to participate in the study came to a centrally located building in the village on the designated survey day. A field team of two investigators from Yale University, one investigator from the University of Michigan and two locally trained field assistants conducted the data collection.
A convenience sample of 319 mother and child pairs participated in the study. Since few data were available on prevalence of the health conditions being measured, the study was powered to detect an 8–10 % difference in the prevalence of obesity compared with the 1999 Samoa National Nutrition Survey (estimate based on an increase in obesity of over 20 % among adults over the same period). Three hundred dyads were needed to detect this difference at 80 % power, α=0·10. Interest in participation exceeded that target, resulting in a sample size of 319 dyads.
Eligible children were 24–59 months old and of Samoan origin (based on having four Samoan grandparents). Eligible mothers were not pregnant (assessed by self-report), had no severe physical or cognitive impairment (assessed by interviewers), and were willing and able to complete the interview portion of the study. Where mothers had two or more children who met study criteria, we randomly selected one for participation. For the present study, we excluded fourteen children: data checking identified two children under 24 months old, ten lacked complete physical measurements, and two were excluded based on having been born at <37 weeks’ gestation to avoid the sequelae of prematurity biasing analyses of growth data. Therefore, the final sample for analyses included 305 mother and child pairs.
Physical measurements
Anthropometric measurements were taken from both mothers and children, with all participants in light island clothing. Standing height was measured to the nearest 0·1 cm using a portable stadiometer (Pfister Imports, NY, USA) and weight to the nearest 0·1 kg using a Tanita HD 351 digital weight scale (Tanita Corporation of America, IL, USA). Duplicate measures were averaged for use in analyses. A finger-prick blood Hb sample was also taken using an AimStrip Hb test system (Germaine Laboratories Inc., TX, USA).
Child nutritional status outcomes
The 2006 WHO growth standards were used to construct dichotomous indicators of stunting, overweight/obesity and underweight. Children whose height-for-age was below −2 sd, BMI-for-age was above +2 sd or weight-for-age was below −2 sd from the corresponding median of the reference population (height-for-age Z-score (HAZ) <−2, BMI-for-age Z-score (BMIZ) >+2 or weight-for-age Z-score (WAZ) <−2) were classified as stunted, overweight/obese or underweight, respectively( Reference de Onis 17 ). Child anaemia was defined as having Hb level below 110 g/l( 18 ).
Child, maternal and household characteristics
A questionnaire, administered in Samoan to all mothers, collected detailed information on child, maternal and household characteristics.
Child characteristics included health history, infant feeding practices (breast-feeding and medication use), nutrition and maternal-reported physical activity. We used a dietary screener to assess general dietary behaviours and frequency, including eating breakfast daily and consuming specific types of foods, such as eating frozen desserts on a weekly basis. Dietary intake was assessed using a 104-item FFQ with a 30 d reference period. The questionnaire was based on a previously validated version of the instrument used in adults and school-aged children in Samoa and American Samoa( Reference DiBello, McGarvey and Kraft 14 ) and updated to include four child-specific foods (canned infant foods, puréed fruit/vegetable pouches, infant formula and breast milk). The frequency of intake of individual items was measured across seven categories ranging from never/less than once per month to more than six times per day. We calculated daily total energy intake (TEI) and nutrients by multiplying the daily consumption frequency of each food by the nutrient content of a fixed, standard portion size. The nutrient content of FFQ items was estimated based on the US Department of Agriculture Food Composition Tables( 19 ) and complementary information from the FAO Pacific Island Food Composition Tables( Reference Dignan, Burlingame and Kumar 20 ) for locally produced items. Macronutrient energy was divided by total energy to estimate the percentage of total energy from each macronutrient. Intakes of specific nutrients were adjusted for total energy using the residual method( Reference Willett, Howe and Kushi 21 ).
We used the Netherlands Physical Activity Questionnaire for Young Children (NPAQ)( Reference Janz, Broffitt and Levy 22 ) to assess children’s physical activity level. No locally relevant physical activity questionnaire has been developed for use in the Samoan population; therefore we used a questionnaire designed for use in another developed country setting in the same age group that had been widely used in other settings( Reference Janz, Broffitt and Levy 22 , Reference Bielemann, Reichert and Paniz 23 ). Mothers were asked to describe their child’s level of physical activity and preferences for being active compared with other children using a series of seven 5-point Likert scale questions. Answers were summed to calculate a physical activity score, with a higher score indicating higher levels of activity. The physical activity score was categorized into approximate tertiles for analysis: 0–27 (n 109), 28–29 (n 84) and 30–35 (n 112). In addition, we calculated total screen time by summing the reported minutes spent watching television and using a computer or phone.
We collected data on maternal demographic characteristics, including age, highest level of education attained and marital status. Maternal weight status was classified as either normal (<26 kg/m2) or overweight/obese (≥26 kg/m2) based on Polynesian BMI cut-offs, which are sensitive to the greater lean mass per kilogram in the adult Polynesian population compared with those of other ethnicities( Reference Swinburn, Ley and Carmichael 24 ). Maternal anaemia was defined as Hb level less than 120 g/l( 18 ).
For household characteristics, mothers were asked to select from five category ranges of annual household income, based on the 2011 Samoan Census( 25 ). Based on the distribution of responses, we dichotomized income into less than $10 000 tala v. $10 000 tala or more for the purpose of our analyses. A material lifestyle score (MLS) was calculated for each household based on ownership of eighteen consumer durables (such as television or refrigerator). This type of index is a sensitive measure of family socio-economic status in modernizing societies and has been previously used to study the adult Samoan population( Reference DiBello, McGarvey and Kraft 14 ). We categorized the MLS into approximate quartiles based on the score distribution: 0–2 (n 63), 3–4 (n 68), 5–7 (n 97) and 8–18 (n 77). These quartiles are referred to as low, medium, high and very high MLS, respectively. Mothers’ perception of the community spirit in their village of residence was measured using a 5-point Likert scale question from an adult Samoan study( Reference Hawley, Minster and Weeks 12 ), and categorized as weak, average or strong.
Statistical methods
The analysis for the study proceeded in the following steps. First, we used descriptive statistics to summarize the characteristics of the sample and then performed bivariate analyses to examine sample differences across the regions using the ANOVA F test for continuous variables, the χ 2 test for categorical variables and TEI-adjusted linear regression models for macronutrient and micronutrient intakes.
Second, we examined the unadjusted associations between the sample characteristics and each of the child nutritional status outcomes using χ 2 tests, t tests or TEI-adjusted models, as appropriate.
Finally, multivariable logistic regression analyses were performed to identify independent child, maternal and household correlates of the child nutritional status outcomes. For each outcome, we used a backward elimination strategy to determine the most parsimonious model. Variables significant at the 0·10 level were retained in the final models, given the limited sample size. In addition, we included census region, child’s age group and sex, and TEI in all models regardless of their statistical significance to ensure that observed associations were not confounded by these sample characteristics. All analyses were performed using the statistical software package SAS version 9.4.
Results
Characteristics of the study population
Mean age of the children and mothers was 40·0 (sd 10·2) months and 33·8 (sd 9·6) years, respectively. The number of mother–child pairs was similar across the urban AUA (n 100), peri-urban NWU (n 102) and rural ROU (n 103) regions.
Table 1 presents the characteristics of the children and mothers by census region. The distribution of child age and sex did not differ between census regions. Based on data from the dietary screener, we found significant differences in children’s fruit and vegetable consumption across regions (P<0·01 for both), with the highest levels in the rural ROU region and lowest in the urban AUA region. Rural ROU children were also more active, based on their NPAQ scores, than children in the other regions. Urban AUA children had the highest mean total screen time (90·5 (sd 60·0) min/d). Level of maternal education varied across census regions, with the most high-school graduates in the urban AUA (70·0 %) and the least in the rural ROU region (48·5 %). Urban AUA households were wealthier and had a higher MLS compared with those in the peri-urban NWU and rural ROU regions.
Table 1 Description of the study population by census region; Samoan island of Upolu, June–August 2015
* Percentages may not sum to 100 % due to missing data.
† P values for ANOVA F test (for continuous variables) or χ 2 test (for categorical variables).
‡ Total screen time is the sum of the reported minutes spent watching television, on the computer or on the phone.
§ Physical activity score is the sum of items from the Netherlands Physical Activity Questionnaire for Young Children (NPAQ). The score is out of a maximum of 35, with higher scores indicating greater physical activity.
║ Material lifestyle score is a sum of consumer durables owned (fridge, freezer, stereo, portable stereo, microwave oven, rice cooker, blender, sewing machine, television, VCR/DVD, couch, washing machine, landline telephone, computer/laptop, tablet, electric fan, air conditioner and motor vehicle). The score is out of a maximum of 18, with higher scores indicating higher material lifestyle.
Child daily dietary intake, based on the FFQ, is presented in Table 2. Mean daily TEI was significantly different across the three census regions, with the children living in the rural ROU region having the lowest total energy intake (10 924 kJ/d (2611·0 kcal/d) v. 12 409 kJ/d (2965·8 kcal/d) in AUA and 18 960 kJ/d (4531·6 kcal/d) in NWU; Table 2), although the energy intake reported in all regions far exceeded the caloric recommendations for this age group( 26 ). Sugar intake also differed by census region, with the NWU region consuming the most (159·6 g/d) and the most rural ROU region consuming the least (135·3 g/d). Among the micronutrients, only TEI-adjusted mean daily intakes of vitamin C, Ca, cholesterol and Na differed across regions; daily intakes of vitamin C and Ca were greatest among peri-urban NWU children while mean daily cholesterol and Na intakes were greatest among rural ROU children.
Table 2 Daily energy, macronutrient and micronutrient intakes of child participants by census regionFootnote *; Samoan island of Upolu, June–August 2015
TEI, total energy intake.
* All nutrient intakes adjusted for TEI based on child FFQ data.
† P values for ANOVA F test (for TEI and amount of nutrient intake per day) or TEI-adjusted linear regression (for macronutrient intake as a percentage of TEI).
Associations between child, maternal and household characteristics and child nutritional status
The overall prevalence of stunting, overweight/obesity and anaemia was 20·3 % (n 62), 16·1 % (n 49) and 34·1 % (n 104), respectively. Only seven children were underweight (2·3 %; data not shown) and therefore characteristics associated with underweight are not described in the current paper. Despite the differences in child, maternal and household characteristics described above, census region was not significantly associated with any nutritional status outcome. Tables 3 and 4 show the unadjusted associations between sample characteristics and nutritional status outcomes. Any characteristic with an unadjusted association with the outcomes at a significance level of 0·01 was carried forward into the multivariable models reported in Table 5.
Table 3 Unadjusted associations between sample characteristics and child nutritional status outcomes; Samoan island of Upolu, June–August 2015
HAZ, height-for-age Z-score; OW/OB, overweight and obesity; BMIZ, BMI-for-age Z score.
* Numbers may not sum to totals due to missing data.
† P value for χ 2 test.
Table 4 Child daily dietary intake by nutritional status outcomes (n 305)Footnote *; Samoan island of Upolu, June–August 2015
HAZ, height-for-age Z-score; OW/OB, overweight and obesity; BMIZ, BMI-for-age Z score; TEI, total energy intake.
* Values based on child FFQ data.
† P value for t test (for TEI and amount of nutrient intake per day) or TEI-adjusted linear regression (for macronutrient intake as percentage of TEI).
Table 5 Multivariable logistic regression models of sample characteristics associated with child stunting, overweight/obesity and anaemia; Samoan island of Upolu, June–August 2015
OW/OB, overweight and obesity; TEI, total energy intake; ref., reference category.
* One participant was excluded due to a missing value for maternal anaemia.
Odds of child stunting were approximately 59 % lower for females compared with males, after adjusting for other characteristics (P=0·006). Children who ate frozen desserts at least weekly had 70 % lower odds of stunting compared with those who did not (P<0·001). For every 10 mg increase in daily vitamin C consumption, we found 4 % increased odds of stunting (P=0·046). While eating breakfast daily was associated with stunting in the bivariate analysis, the association was not significant in the multivariable model (OR=2·09; 95 % CI 0·93, 4·68, P=0·073).
Children with very high MLS had nearly fourfold greater odds of overweight/obesity compared with those with low MLS (OR=3·61; 95 % CI 1·22, 10·71, P=0·016). The odds of overweight/obesity decreased by 11 % for every 10 g increase in daily sugar intake (P=0·032).
After adjusting for other characteristics, odds of child anaemia decreased with age. Among children aged 48–59 months, the odds were 75 % lower than among children who were 24–35 months old (P<0·001). For every 100 mg increase in daily Ca intake, the odds of anaemia increased by 14 % (P=0·052). Children with a married mother had a lower odds of anaemia (OR=0·60; 95 % CI 0·34, 1·07) than children with single mothers, although the association was not significant (P=0·084). Children with an anaemic mother had more than twice the odds of anaemia compared with children whose mother’s Hb level was within the healthy range (OR=2·20; 95 % CI 1·22, 3·98, P=0·007).
Dual burden of malnutrition
Although levels of stunting and anaemia did not differ significantly between children who were overweight/obese and those who were not (P=0·118 and P=0·175, respectively), among the forty-nine overweight and obese children, 28·6 % (n 14) were also classified as stunted and 42·9 % (n 21) were anaemic (data not shown). Of the sixty-two children who were stunted, 48·4 % (n 30) were anaemic; however, this proportion of anaemia was significantly lower than in those who were not stunted (51·6 %, n 32; P=0·008).
Discussion
Adequate nutritional status is critical for child growth and development and health later in life. The present study identified a worrisome prevalence of malnutrition among young Samoan children, which needs to be addressed with targeted intervention. While child, maternal and household characteristics differed by census region, there were no differences in the prevalence of stunting, overweight/obesity or anaemia by census region, indicating that, despite differential exposure to urbanization and the nutrition transition, all Samoan children are at risk of poor nutritional outcomes. Several child, maternal and household characteristics were identified to be associated with each nutritional status outcome, giving suggestions for intervention targets.
The prevalence of stunting (20·3 %), overweight/obesity (16·1 %) and anaemia (34·1 %) observed in the current study was substantially higher than reported in the 1999 Samoan National Nutrition Survey. While the two studies are not directly comparable, as the National Nutrition Survey included children living on the island of Savai’i, which is the most rural area of Samoa, the increased prevalence of malnutrition reported in the current study suggests that the development, modernization and nutrition transition that have taken place in Samoa over the past 16 years has not benefited child nutritional status. This is in line with other settings in which urbanization and nutritional transition have had detrimental effects on child health( Reference Popkin 2 ) because it exposes children to diets of higher energy content, but lower quality( Reference Drewnowski and Popkin 4 ) and an increased burden of infectious disease associated with lower income and more crowded, urban housing( Reference Alirol, Getaz and Stoll 27 ).
Stunting reflects chronic undernutrition during the most critical periods of growth and development in early life. Overall, the prevalence of stunting observed in the present study was consistent with reports from other Pacific Island nations: 48 % of children under 5 years of age are reportedly stunted in Papua New Guinea, 33 % in the Solomon Islands and 26·3 % in Vanuatu( Reference Grieve, Busch-Hallen and Mellor 28 ). The majority of the Samoan children who were classified as stunted were in the youngest age group of 24–35·99 months, while there was no stunting in the oldest age group. Although these data are cross-sectional, they suggest that Samoan children may regain position on the WHO height growth charts as they get older. This requires confirmation with longitudinal data, and further information on the potential consequences of this pattern of slower height growth for later health and cognitive development is also needed.
Young girls in the current study had decreased odds of stunting. This finding is consistent with studies in many other settings globally( Reference Wamani, Astrom and Peterson 29 – Reference Foster, Byron and Reyes-Garcia 31 ) and across the Pacific, including Papua New Guinea( Reference Decaro, Decaro and Worthman 32 ) and Vanuatu( Reference Grieve, Busch-Hallen and Mellor 28 ). Our data did not suggest that this was a result of different infant feeding practices or current nutritional status, and there is little evidence from the Pacific to suggest that daughters are preferred( Reference Marshall 33 ), although this has not been exhaustively explored in contemporary populations. It has, however, been argued that improved growth in female compared with male children may reflect their relative resilience to adverse in utero environments( Reference Stinson 34 ), seasonal fluctuations in nutritional availability and infectious diseases( Reference Decaro, Decaro and Worthman 32 ).
Eating frozen desserts (such as ice cream) at least weekly was also found to be associated with decreased odds of child stunting; however, we suspect that this may reflect higher family socio-economic status and the child’s overall energy intake, rather than specific nutritional content obtained from the desserts. In line with this hypothesis, consuming frozen desserts at least weekly was associated with higher MLS and mean daily TEI (data not shown). Contrary to the existing literature( Reference Simeon and Grantham-McGregor 35 – Reference Black, Allen and Bhutta 37 ), stunting was positively associated with the child eating breakfast daily (significant in the bivariate model and P=0·07 in the multivariable model) and daily intake of vitamin C. These findings are difficult to explain and warrant further investigation. There are inherent limitations of dietary intake estimation based on FFQ in children( Reference Burrows, Martin and Collins 38 , Reference Roman-Vinas, Ortiz-Andrellucchi and Mendez 39 ), which may have contributed to these findings, particularly as we were not able to determine the dietary composition of breakfast specifically. Additionally, we were not able to account for reverse causality. Physicians may have already detected or parents may have recognized the low height-for-age among the stunted children in our sample and, in response, parents may have increased the child’s dietary intake.
Looking at the opposite end of the spectrum of malnutrition, we found problematic levels of childhood overweight/obesity. This is especially concerning since overweight children are known to be at increased risk of becoming overweight adolescents and adults( Reference Ng, Fleming and Robinson 40 ) and of experiencing metabolic dysregulation in later life( Reference Anderson 6 ). Estimates of overweight/obesity among young children in Pacific Island nations are difficult to compare as different classifications for weight status, such as weight-for-height, weight-for-age or BMI-for-age( Reference de Onis 17 ), have been used. However, in American Samoa, which is further along in the nutrition transition, it is estimated that approximately 35 % of children under 5 years of age are overweight/obese using comparable WHO BMI-for-age Z-scores( 41 ). In the Commonwealth of the Northern Mariana Islands, childhood obesity was estimated to be approximately 25 % in 2–3-year-olds and 26 % in 4–6-year-olds( Reference Paulino, Coleman and Davison 42 ). The strongest correlate of overweight/obesity in our sample was household MLS. High family socio-economic status, indicated by higher MLS score, increased the odds of overweight/obesity among young children, a pattern similar to that found among adults in this setting( Reference Keighley, McGarvey and Quested 10 , Reference Seiden, Hawley and Schulz 13 , Reference DiBello, McGarvey and Kraft 14 ).
High BMI-for-age often results from a chronic positive energy balance( 5 , Reference de Onis 17 ); however, we did not find any differences in either maternal-reported physical activity level or mean daily TEI between young Samoan children who were overweight/obese and those who were not. Furthermore, there were no differences in the proportion of energy consumed from macronutrients (protein, carbohydrates and total fat) between the two groups. Interestingly, TEI-adjusted Ca intake was associated with increased odds of overweight/obesity in our bivariate analyses, consistent with the suggestion that dietary Ca intake may positively influence energy balance, particularly in overweight/obese children, although it did not reach the threshold of significance in our multivariable model( Reference Astrup 43 ). Overall, dietary sugar intake was very high among children in our sample. In addition, we found an inverse association between sugar intake and overweight/obesity; however, this finding should be interpreted with caution, as we cannot differentiate between added v. natural sugars in the foods consumed by the child. Local fruits and vegetables, and starchy staples in the traditional Pacific Island diets, are very carbohydrate-dense( Reference Dignan, Burlingame and Kumar 20 ), and these natural sugars contribute significantly to overall daily sugar intake. Commercial, processed food and beverage products may be nutritionally detrimental, potentially increasing dietary consumption of added sugar and displacing consumption of other more nutritious options in early years of life( Reference Seiden, Hawley and Schulz 13 , Reference Pries, Huffman and Adhikary 44 ). As may be the case with stunting, there could be issues of reverse causality if parents had recognized overweight and implemented dietary changes.
We found that more than one-third of the study sample (children and mothers) had some degree of anaemia. We included mild anaemia in our outcome, as the WHO cautions that even mild levels of anaemia may be problematic for growth and development in children under 5 years of age( 18 ). The prevalence of anaemia among the Samoan children was consistent with estimates from other countries across the Pacific region. In the Solomon Islands, 48 % of children aged 6–59 months are anaemic( Reference Grieve, Busch-Hallen and Mellor 28 ), while more than half of children of the same age group are anaemic in Nauru( 45 ) and Tuvalu( Reference Grieve, Busch-Hallen and Mellor 28 ). In neighbouring American Samoa, the prevalence of anaemia is very similar in this age group: 33 % among those aged 1–4 years( Reference Kemmer, Novotny and Ah Ping 46 ). Also similar is the relationship between anaemia and age, with the odds of anaemia decreasing with age in both settings( Reference Kemmer, Novotny and Ah Ping 46 ). The odds of any anaemia were nearly 120 % higher among children with an anaemic mother compared with those without, as expected based on the strong evidence linking maternal and child nutritional status( Reference Tzioumis and Adair 47 ). Having a married mother was protective against child anaemia, and this supports the link between good health and nutrition behaviours in children with social support provided through marriage( Reference Schafer, Schafer and Dunbar 48 , Reference Combs-Orme, Wilson and Cain 49 ). Increasing Ca intake increased the odds of child anaemia (although the association was just above P<0·05 (P=0·052) in the final model); this finding may reflect the competition between Ca and Fe for absorption in the body, which can adversely impact Hb levels, although, again, we acknowledge the limitations of the FFQ in measuring child dietary intake( Reference Bailey, West and Black 50 ).
While Fe deficiency is the most common nutritional source of anaemia in young children( Reference Lopez, Cacoub and Macdougall 51 ), the current study did not find an association between the two, suggesting that perhaps the anaemia has an infectious cause. Multiple mosquito-borne illnesses and helminth infections, such as hookworms, can result in anaemia( Reference Kassebaum 52 ) and further investigation is urgently needed to determine the source of anaemia among these young children. A recent study suggested that hookworms were present in the faeces of almost all dogs in Samoa (97·5 %)( Reference Sjölander 53 ); with a large and increasing population of stray dogs, this may represent a major public health concern.
The finding of coexisting levels of child stunting, overweight/obesity and anaemia highlights the vulnerability of young Samoan children to malnutrition in various forms, simultaneously. As in other low- and middle-income countries undergoing the nutrition transition, it is common to find a dual burden of malnutrition: undernourishment and micronutrient deficiencies, coexisting with the other extreme of overweight/obesity and other diet-related chronic diseases( Reference Eckhardt 54 , Reference Doak, Adair and Bentley 55 ). Obesity and micronutrient deficiencies are each problematic, but in combination exacerbate the severity of nutrition-related NCD( Reference Garcia, Long and Rosado 56 ). Previous studies have consistently shown an association between childhood stunting and anaemia( Reference Tzioumis and Adair 47 , Reference Bailey, West and Black 50 , Reference Lopez, Cacoub and Macdougall 51 ). Especially among the poor and those living in environments with insufficient sanitation, children are at higher risk of infection and micronutrient deficiencies( Reference Bailey, West and Black 50 ). While the global prevalence of undernutrition decreased between 1990 and 2012, the number of countries with equivalent levels of stunting and overweight among children under 5 years has increased substantially( Reference Tzioumis, Kay and Bentley 57 ).
Surprisingly, no common correlates of child nutritional status outcomes were observed in the present study. This finding suggests that different intervention approaches may be necessary to prevent malnutrition among young children in this setting. The study identified unique child, maternal and household characteristics associated with child nutritional status, which can potentially help us to identify those at risk of poor nutritional outcomes and target interventions( Reference Hawkes, Haddad and Udomkesmalee 58 ). Further study is needed to fully understand the origins of stunting and anaemia in this setting, including a comprehensive assessment of the possible infectious causes. However, overweight/obesity prevention efforts are clearly needed for this age group and can begin in the households at high risk.
The current study had some limitations. The findings may not be generalizable, as the study purposely recruited similar numbers of mother–child pairs from each of the three census regions and was therefore not nationally representative. The demographics, community layout and urban–rural setting on the Samoan island of Upolu may make any associations specific to this area in particular. As a result of our convenience sampling, there is the possibility that the prevalence estimates of stunting, overweight/obesity, underweight and/or anaemia may have been either under- or over-represented. Comparing our study sample with the 2014 Samoan Demographic and Health Survey showed that our participants were less educated than the general population (60·3 % had secondary education compared with 76 % of the general population) and were of slightly lower socio-economic status based on household ownership of consumer durables( 59 ). Additionally, due to the cross-sectional approach of the study, we are unable infer causality and tease apart temporality to determine potential prospective risk factors for the child nutritional status outcomes; further longitudinal studies are needed to do this.
To our knowledge, the present study is the first regional study to report child, maternal and household characteristics associated with child nutritional status outcomes in children aged 2–5 years. Poor nutrition at this age has lifelong implications, reducing a child’s ability to learn and grow to her/his full potential, and can lead to decreased productivity and weaker long-term national economic performance( 5 , Reference Hawkes, Haddad and Udomkesmalee 58 ). Nutrition-related NCD control and prevention strategies that target undernutrition and overweight/obesity simultaneously in this age group are urgently needed in Samoa and potentially elsewhere in the Pacific Islands. An investment in nutrition is foundational for human development and will help to ensure good health throughout the life course.
Acknowledgements
Acknowledgements: The authors would like to thank the study participants. They are grateful to the Samoan Ministry of Health (especially Christina Soti-Ulberg and staff at the Office of Nutrition Surveillance), Samoa Bureau of Statistics, and Ministry of Women, Community and Social Development for their support and facilitation of the study. Financial support: This work was supported by the David Dull Internship Fund (Yale University; CCC); the Stolwijk Fellowship (Yale University; J.J.P.); and Yale University and University of Michigan Faculty Funds. The funders had no role in the design, analysis or writing of this article. Conflict of interest: None. Authorship: C.C.C. and N.L.H. conceived the study and, with M.M.D., wrote the initial draft of the manuscript. C.C.C., J.J.P. and E.A.F. conducted the research with the support of T.N., M.S.R., N.C.D. and R.L.D. C.C.C., M.M.D., A.A.T. and N.L.H. analysed and interpreted the data. All authors read and approved the final manuscript. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Yale University Institutional Review Board and the Health Research Committee of the Samoan Ministry of Health. Written informed consent was obtained from all participants.
