Nutrient Intake, Diet Quality, and Weight Measures in Breakfast Patterns Consumed by Children Compared with Breakfast Skippers: NHANES 2001–2008

Most studies showing that children consuming breakfast have better nutrient intakes, diet quality, and lower weight than breakfast skippers have the incorrect premise that breakfast meals are homogeneous. The purpose of this study was to classify breakfast meals into patterns and determine the association of the breakfast patterns with daily and breakfast nutrient intakes, diet quality, and weight. Data from children (2–18 years of age; N = 14,200) participating in the National Health and Nutrition Examination Survey 2001–2008 were used. Intake was determined from one day 24-hour dietary recalls. Diet quality was measured using the Healthy Eating Index-2005 (HEI-2005). Body mass index (BMI) z-scores were determined. Twelve patterns (including No Breakfast [∼19% of population]), explaining 63% of the variance in energy from breakfast, were examined. Covariate adjusted general linear models were used to compare outcome variables of consumers of different patterns with breakfast skippers. The p value was Bonferroni corrected (< 0.05/12 = < 0.0042). Consumers of the Eggs/Grain/Meat, Poultry, Fish (MPF)/ Fruit Juice (FJ) and MPF/ Grain/FJ patterns showed higher daily intakes of saturated fats, solid fats, and sodium and lower daily intakes of added sugars than breakfast skippers. Consumers of most breakfast patterns showed higher daily intakes of some nutrients of public health concern (dietary fiber, vitamin D, calcium, and potassium); however, those consuming the Grain or MPF/Grain/FJ pattern did not. Consumers of the Grain/Lower Fat Milk (LFM)/Sweets/FJ, Presweetened (PS) Ready-to-eat Cereal (RTEC)/ LFM, RTEC/LFM, Cooked Cereal/Milk/FJ, and Whole Fruit patterns had higher total HEI-2005 scores than breakfast skippers; those consuming the MPF/ Grain/FJ pattern had lower diet quality than breakfast skippers. Consumption of the Grain/ LFM/Sweets/FJ, PSRTEC/whole milk, Soft Drinks/ FJ/Grain/Potatoes, RTEC/whole milk, and Cooked Cereal/ Milk/ FJ patterns was associated with lower BMI z-scores than seen in breakfast skippers. There are dietary and weight advantages of consuming breakfast, especially breakfasts that include grains, cereals, LFM, and fruit/ FJ, in contrast to the potential adverse effects of skipping breakfast.

Breakfast consumers have higher intakes of many nutrients, especially micronutrients [3][4][5][6]18,19], when compared with non-consumers. An early study showed that a higher proportion of children skipping breakfast failed to meet two-thirds of the Recommended Dietary Allowance for vitamins and minerals as compared to breakfast consumers [20]. Dubois, et al. [21], however, showed that Canadian preschool children who skipped breakfast had the same energy, carbohydrate, and fat intake as those children consuming breakfast; however, protein intake was lower in breakfast skippers. Micronutrient intake was not examined in that study. Thus, the association between breakfast consumption and nutrient intake clearly depends on the type of foods consumed and what is consumed during the remainder of the day. When ready-to-eat cereals (RTEC) were consumed as part of the breakfast meal, there was a difference in nutrient intake compared with breakfast skippers or those consuming "other breakfasts" [3]. In addition to better micronutrient intakes, consumption of RTEC at breakfast has been associated with lower daily intakes of fat, saturated fatty acids (SFA), and cholesterol [3] compared to consumption of "other breakfast meals" or skipping breakfast. A higher consumption of milk has also been associated with an RTEC breakfast [22], suggesting that breakfast consumption is associated with consumption of nutrient-dense foods. A limitation of most of those studies is that breakfast and, especially, "other breakfast types" were generally not well defined by the authors and the assumption may have been made that they were homogeneous enough to include in a single group for statistical comparison.
In some studies, breakfast consumption has also been associated with lower body mass indices and other measures of adiposity in children [1,3,4,[6][7][8][9][10][11]. Ready-to-eat cereals [3,[23][24][25][26], including pre-sweetened RTEC (PSRTEC) [27][28][29], breakfasts have specifically been associated with lower measures of weight and adiposity. One study [3] also compared "other breakfasts" with breakfast skipping and RTEC breakfasts with weight/ adiposity parameters and showed that breakfast skippers had higher BMI z-scores and waist circumferences than those consuming RTEC or "other breakfasts." A recent study in adults has shown that individuals consume multiple "breakfast types" or patterns, which provide variable nutrient contributions to the overall diet and to diet quality, and have different associations with weight [30]. That study showed adults consuming containing Grain/fruit juice (FJ); PSRTEC/lower fat milk (LFM); RTEC/LFM/ Fruit/ FJ; or Cooked Cereal had lower BMIs and Waist Circumferences than breakfast skippers, but that no other patterns were associated with different measures of weight or adiposity. It is presumed that children would also show specific breakfast meal patterns with varying effects on nutrient intake, diet quality, and weight, although this has not been studied. The purposes of this study were to identify breakfast patterns consumed by a nationally representative sample of children, to examine nutrient intake resulting from consumption of breakfast patterns, and to determine the association of the breakfast patterns to total daily nutrient intake and diet quality of consumers of specific patterns vs breakfast skippers (No Breakfast pattern). Weight and adiposity measures of those consuming different breakfast patterns were also compared with breakfast skippers.

Study Population
An overview of the National Health and Nutrition Examination Survey (NHANES), including the purpose, study population, sampling strategy, interview and physical examination procedures and response rates have been published on line [31][32][33][34]. In this study, data from children two to 18 years participating in the NHANES 2001-2008 were concatenated to increase sample size [35]. Those with unreliable dietary records, as judged by National Center for Health Statistics staff (n = 359), pregnant or lactating females (n = 117), and children who were breastfeeding (n = 13) were excluded from the study. The final analytic sample consisted of 14,200 children. Stringent protocols and procedures enforced by NHANES ensure confidentiality and protect individual participants [36]. As this was a secondary data analysis which lacked personal identifiers, this study did not require further institutional review [37].

Demographics and Dietary Information
Demographic information was determined from the NHANES interview [38]. Intake data were obtained from in-person 24-hour dietary recall interviews using an automated multiple-pass method [39,40]. Parents/guardians of children two to five years provided the 24 -hour dietary recalls; children (6 to 11 years) were assisted by an adult; and older children provided their own recall. In 2001-2002, a single 24-hour dietary recall was collected in person; however, defined as a BMI ≥ 95 th percentile [53].

Statistical Analyses
Breakfast consumption patterns were identified using SAS 9.2 (SAS Institute, Cary, NC; 2009) PROC CLUSTER. For each participant the percent of energy at breakfast from each of the food groupings [30] was determined. The patterns were identified by percent energy within each food grouping consumed at breakfast at the centroid of each cluster. Initially, eight, 12, and 16 patterns were evaluated; however, for subsequent analyses, the 12 pattern output (which included No Breakfast) was used since this set of breakfast patterns allowed delineation of type of RTEC and the use of LFM. Further, the 12 pattern output explained moderately more variance in energy from the breakfast meal (63% of the variance) than eight patterns (53% of the variance), but not substantially more than 16 patterns (70%). Based on their dietary intake, participants were placed into one breakfast pattern and dietary day 1 weights were used for all analyses [35].
Least-square means ±SE were calculated using PROC REGRESS of SUDAAN for dietary intake and diet quality (HEI-2005) for participants consuming each breakfast pattern. After confirming a significant overall F test for breakfast patterns (entered as categorical variables [patterns [1][2][3][4][5][6][7][8][9][10][11][12], differences for variables of interest were determined via t-test compared to the No Breakfast pattern [referred to as breakfast skippers]). Covariates included: self-reported age, gender, race/ethnicity, and the poverty income ratio (PIR), grouped into three categories (< 1.25, 1.25-3.49, and > 3.49) also served as a covariate. These values are related to the federally set poverty lines, so a PIR of < 1.25 equates to below 125% of the poverty line. Higher values mean the individuals had higher incomes. Physical activity, another covariate, was determined using a questionnaire [54]. Finally, self-reported energy intake for nutrient related variables (not for energy intake itself) was used as a covariate. The HEI-2005 was not controlled for energy intake, since the calculation of the index score controls for energy [49]. Logistic regression was used to assess the likelihood of being overweight, obese, overweight or obese for consumers of each breakfast pattern compared with breakfast skippers. For all analyses, a probability of p < 0.05 was considered significant; however, a Bonferroni correction was applied for multiple comparisons (p < 0.05/12), so the effective p value was p < 0.0042.

Breakfast Clusters and Population Demographics
The 12 breakfast patterns identified were: 1) Grain/LFM/ Sweets/FJ (n = 2,797; 21.8% of the population); 2) No Breakfast (n = 3,018; 18.7%); 3) PSRTEC/LFM (n = 1,684; 15.2%); 4) (Grain n = 1,090; 8.2%; 5) Eggs/Grain/Meat/Poultry/Fish [MPF]/FJ (n=1,068; 6.7%); 6) PSRTEC/Whole Milk (n = 1,145; 6.5%); 7) RTEC/LFM (n = 732; 6.1%); 8) Soft Drink/FJ/Grain/Potatoes (n = 889; 5.9); 9) RTEC/Whole Milk (n = 714; 4.1%); 10) MPF/Grain/FJ (n = 617; 3.5%); 11) Cooked Cereal/Milk/FJ (n = 354; 2.5%); and 12) Whole Fruit (n = 112; 0.8%) ( Table 1). The patterns varied widely by foods consumed and energy contribution of individual foods (Supplemental Tables 1 and 2, respectively). Demographics of the population (N = 14,220) by breakfast cluster are shown in Table 2.   Other (%) 6.9 ± 0.9 6.4 ± 1.1 5.0 ± 0.9 6.0 ± 1.2 6.6 ± 1.5 6.5 ± 1.     49), current physical activity level (sedentary, moderate and vigorous), alcohol intake (g/d), and energy  = significantly different from no breakfast at p< 0.0042 Abbreviations: LSM = least square mean; SE = standard error; Gr = grain; LFM = lower fat milk; FJ = fruit juice; PSRTEC = pre-sweetened ready-to-eat cereal; MPF = meat, poultry, fish; RTEC = ready-to-eat cereal Table 3 presents the absolute intake of energy, protein, and nutrients to limit for the breakfast meal only and for the entire day. To help the reader understand the contribution of the breakfast meal to the entire day's intake, the energy and nutrients have been converted to percentages in this section. The percent energy contributed by the breakfast meal to the daily total varied widely among those consuming different breakfast patterns. The Whole Fruit breakfast pattern was associated with the lowest (6%) and MPF/Gr/FJ breakfast pattern was associated with the highest (25%) percent of daily energy intake. A Whole Fruit breakfast also contributed the lowest percent protein (2%), added sugars (3.6%), SFA (0.3%), solid fat (3%), cholesterol (3%) and sodium (2%) to the daily intake of these nutrients. The percent of protein contributed by the breakfast meal to the daily total was highest among consumers of the Eggs/Grain/MPF/FJ breakfast pattern (30%). A Soft Drink/FJ/Grain/Potatoes breakfast contributed the highest percent intake of added sugars (24%) to the daily intake. An Eggs/Grain/MPF/FJ breakfast contributed the highest (31%) percent total daily intake of SFA and cholesterol (73%). The MPF/Gr/FJ breakfast pattern contributed the highest (36%) percent total daily intake of solid fat and sodium (32%). Table 4 presents the absolute intake of nutrients of public health concern and shortfall nutrients for the breakfast meal only and for the entire day. To help the reader understand the contribution of the breakfast meal to the entire day's intake, the energy and nutrients have been converted to percentages in this section. For the breakfast meal only, consumers of all breakfast patterns had higher intakes of all nutrients examined than breakfast skippers, except for those consuming the Whole Fruit breakfast patternthey had a vitamin A intake which was not different from breakfast skippers. The Cooked Cereal/Milk/FJ breakfast contributed the highest (29%) and the Eggs/Grain/MPF/FJ breakfast contributed the lowest (13%) percent fiber to the daily intake of dietary fiber. Except for the Cooked Cereal/Milk/FJ pattern, only the breakfast patterns with milk (PSRTEC/LFM; PSRTEC/Whole Milk; RTEC/LFM; RTEC/Whole Milk) contributed high percentages of vitamin D to the daily intake. The Whole Fruit pattern contributed the lowest percentage (0.6%) of calcium to the total daily intake; and the Grain pattern contributed the lowest percentage of potassium (11%) to the total daily intake.

Energy, Protein, Nutrient Intake, and Diet Quality
For the entire day, energy intake ranged from approximately 1630 kcals/day for those consuming the Whole Fruit breakfast pattern to 2270 kcals/day for those consuming the MPF/Grain/FJ breakfast pattern (Table 3). For those in all patterns, except those in the Whole Fruit breakfast pattern, higher daily energy intakes were seen compared to breakfast skippers. Mean daily protein intake for those consumers of all breakfast patterns, except Grains and Whole Fruit, was higher than that seen among breakfast skippers. For those in the 12 breakfast patterns, daily intake of added sugars ranged from approximately 19-24 tsp/day. For those in the Grain/LFM/Sweets/FJ, Eggs/Grain/MPF/FJ, and RTEC/Whole Milk breakfast patterns, daily intake of added sugars was lower than that seen among breakfast skippers. Only the consumers of the PSRTEC/LFM breakfast pattern had a lower daily intake of SFA than breakfast skippers; whereas, consumers of the Eggs/Grain/MPF/FJ, RTEC/Whole Milk, and MPF/Grain/FJ breakfast patterns had a higher daily intake of SFA than breakfast skippers. Only consumers of the Eggs/Grain/MPF/FJ and the MPF/Grain/FJ breakfast patterns had higher in total sodium intakes than breakfast skippers.
When total mean daily intake of dietary fiber was considered, those consuming the PSRTEC/LFM, RTEC/LFM, Cooked Cereal/Milk/FJ, and Whole Fruit breakfast patterns consumed more dietary fiber than breakfast skippers (Table 4); however, mean intake of consumers of all breakfast patterns was low. When mean daily Vitamin D intake by consumers of the breakfast patterns were compared with breakfast skippers, only those placed in the Grain, Soft Drinks/FJ/Grain/potatoes, MPF/Grain /FJ, and Whole Fruit breakfast patterns did not have higher intakes of Vitamin D intake than breakfast skippers. Consumers in all breakfast patterns except Grain, MPF/Grain/FJ, and Whole Fruit had higher mean daily intakes of calcium than breakfast skippers. Consumers in all breakfast patterns except Grain (lower than breakfast skippers) and MPF/Grain/FJ (not different from breakfast skippers) had higher intakes of potassium than breakfast skippers. Also presented in Table 4 are daily intakes of shortfall nutrients, by those in the different breakfast patterns compared with breakfast skippers. Only consumers in the Grain, MPF/Grain/FJ, and Whole Fruit breakfast patterns did not have daily intakes of vitamin A that were higher than breakfast skippers; whereas, only consumers in the Soft Drink/FJ/Grain/Potatoes breakfast pattern had a higher mean intake of vitamin C than breakfast skippers. For folate, iron, and magnesium, consumers of the Gr/LFM/sweets/FJ, PSRTEC/LFM, PSRTEC/Whole Milk, RTEC/LFM, RTEC/Whole Milk, and Cooked Cereal/Milk/FJ breakfast patterns had higher intakes than breakfast skippers. Finally, consumers of the Whole Fruit breakfast pattern also had higher intakes of magnesium, when compared with breakfast skippers.  Table 5 shows weight (kg), BMI, BMI z-score, and the percentage of the population that were overweight, obese, or overweight or obese by breakfast pattern. Mean BMI z-scores were lower among consumers of five of the breakfast patterns,  Covariates: age, gender, race/ethnicity, poverty income ratio, physical activity, and alcohol intake.

Discussion
A novel contribution of this study was the identification of 12 distinct breakfast patterns and their differential association with nutrient intakes, diet quality, and weight in children. This study showed that nearly 19% of the population skipped breakfast (No Breakfast Pattern); 11 specific breakfast patterns consumed by children were also identified. Mean nutrient consumption among the consumers assigned to the different patterns varied both for the breakfast meal and for the day; not all patterns differed significantly from breakfast skippers. Diet quality also varied among the breakfast consumption patterns. Only consumers of breakfast patterns characterized by cereal or grain intake (Gr/LFM/Sweets/FJ; PSRTEC/whole milk; RTEC/whole milk; Cooked Cereal/Milk/FJ), ironically except the actual "Grains," pattern, which included primarily sweetened grains such as doughnuts, cinnamon buns, and muffins and the RTEC with LFM pattern, had lower mean BMI values than breakfast skippers.
Breakfast consumers have been shown repeatedly to have higher daily intakes of vitamins and minerals, than non-consumers [1-6, 19, 20, 22, 21, 23-26]. The previous literature suggests that children who skip breakfast may not make up for missed nutrients during the day [20], underscoring the importance of the breakfast meal. However, our study suggested this was true for consumers of some breakfast patterns, but not all. Consumers of the majority of breakfast patterns had higher daily intakes of most nutrients of public health concern, as defined by the 2010 DGA [48], as well as some potentially underconsumed nutrients by some sub-populations [48] when compared with breakfast skippers. The varying results with different breakfast patters suggested that the composition of the breakfast meal is important. This finding was not surprising since the foods contributing the highest levels of energy in the Grain pattern were breakfast tarts, doughnuts, pancakes and waffles, pizza, and noodle soups. The foods contributing the highest levels of energy in the MPF/Grain/FJ pattern were sausage biscuits, pork sausage, whole milk, French fries, bacon, and sausage and full-fat cheese on English muffins. These foods tend to be energy dense and may contribute high levels of nutrients to limit, such as SFA and sodium.
In this study, the finding that children who consumed RTEC, including PSRTEC, breakfast patterns had high intakes of most micronutrients of concern confirmed results from other studies [3,22,23,[25][26][27]. Virtually all RTEC are vitamin and mineral fortified, so these results were not surprising. Further, consumption of RTEC has been shown to increase consumption of fluid milk [22], which contributes vitamins A and D, calcium, and potassium to the diet.
None of the breakfast patterns showed a mean daily intake of added sugars that exceeded the Institute of Medicine's threshold of 25% of energy [55]. This included the patterns with PSRTEC, soft drinks, and sweets suggesting a moderate intake of these foods or consumption of foods lower in added sugars throughout the day. It should be noted however, that the mean intake of added sugars ranged from 13% (MPF/Grain/FJ) to 21% (breakfast skippers). Consumers of all patterns, including the No Breakfast pattern, exceeded the total daily recommendations for percent energy from SFA [55] and for sodium [56], suggesting that even those children consuming breakfast patterns characterized by foods low in these nutrients, such as those patterns including LFM, RTEC or cooked cereal, and fruit or FJ, need to reduce intake of foods high in SFA and sodium at other meals and snacks throughout the day.
Results from this study suggested that simply consuming breakfast was not associated with a higher diet quality when compared with breakfast skippers, but that the specific foods or food groups consumed at the breakfast meal may have influenced total diet quality. It is also possible that those consuming foods generally regarded as "healthy" at the breakfast meal have an overall "healthier" eating pattern throughout the day; although overall, children had poor diet quality, regardless of the breakfast pattern consumed. Further studies are needed to confirm this. There is no standard definition of a high quality or nutrient-dense breakfast, despite the recommendation from the 2010 DGA to consume a "nutrient-dense" breakfast. The 2010 DGA provide no recommendations for consumption of specific nutrients or foods to be consumed at breakfast [48]. There have been some attempts have been made to define a "high-quality" breakfast. A study of adolescents [57] used quantitative and qualitative aspects of breakfast foods, the frequency of breakfast consumption, and the energy contribution of the breakfast meal to designate a breakfast score for defining three groups: no breakfast, or a "low quality" breakfast, or a "good/excellent quality" breakfast. Those consuming "good quality" breakfasts had higher intakes of bread, fruit, vegetables, milk, and FJ and lower intakes of soft drinks than those consuming "low quality breakfasts" [57]. Another study defined "high-quality breakfasts" as those including whole grain, fruit or FJ, and low-fat milk products or other sources of calcium [58].
In this study, children consuming breakfast patterns with significantly higher overall diet quality, when compared with breakfast skippers were generally those patterns that included grains, LFM, and fruit or FJ. In adults, it has been shown that good diet quality is essential to lowering the risk of all-cause mortality, cardiovascular disease, and cancer [59][60][61]. One would assume that the same reduction of cardiovascular disease risk would be true in children, especially adolescents; however, this has not been well studied. One study showed that scores on the dietary guideline index for children and adolescents were associated with nutrient intake, but not adiposity [62]. In a study of adolescents, a "healthy diet" coupled with physical activity has been shown to be associated with a decreased cardiovascular disease risk [63]. Finally, Papoutsou, et al. [5] showed that females, but not males, who consumed breakfast, had lower levels of cardiovascular risk factors than breakfast skippers.
Most [1, 3-9, 19, 21, 22-25], but not all [64], studies have shown that consumption of breakfast has been associated with lower weight parameters in children. Our study showed that those consuming some, but not all, of the breakfast patterns in this study had lower BMI z-scores than breakfast skippers. This is consistent with others who have made initial attempts to characterize the type of breakfast consumed with weight or adiposity parameters. Deshmukh-Taskar, et al. [3] in a study of children 9-18 years with breakfast skippers, and RTEC or "other breakfast" consumers the breakfast skippers had the highest weight, followed by those consuming "other breakfasts," and then by RTEC consumers. Ferná ndez Morales, et al. also showed an inverse association between BMI and consumption of calcium, fiber, dairy products, and cereal; they concluded that weight was related to "breakfast quality" [65]. Cho, et al. [66] found similar results, using NHANES III data, although their study also included adults.
Deshmukh-Taskar, et al. [3] suggested multiple reasons for the association between breakfast consumption and lower weight, including rebound overeating at other meals, diet induced thermogenesis, or consumption of more evenly distributed energy intake throughout the day. Although these reasons are plausible, the amount of energy and potentially specific foods consumed at breakfast and throughout the day clearly influence the association between breakfast consumption and weight and these associations have not been well studied.
Examining the consumption of the individual breakfast patterns with the daily total nutrient intake was of particular interest. Consumption of a "Whole Fruit" breakfast pattern gives the impression it would be inherently "healthy" and associated with an overall positive nutrient intake and weight status; however, this pattern contributed little to daily nutrient intake among consumers and had a prevalence of overweight/obese of 36%, which was not different from that seen in breakfast skippers. These findings also suggest that further studies are needed to determine associations of consumption of different breakfast patterns or breakfast skipping with gender, behavioral considerations, or socioeconomic status; the source of the breakfast meal (i.e. at home, at school, or at fast food restaurants); the association of specific breakfast foods/meals and subsequent intake throughout the day; and barriers to breakfast consumption in children. In this way, it may be able to determine more fully the importance of breakfast and its relationship to health.
Study Strengths and Limitations: NHANES is the largest available nationally representative surveillance program in the US and allows use of a very large sample size. Use of pattern analysis to determine breakfast clusters is a novel approach to the study of the breakfast meal. Limitations of this study are that NHANES data are cross-sectional; thus, cause and effect cannot be inferred. A recent meta-analysis cautioned about over-interpretation of results from cross-sectional studies that have examined breakfast consumption and weight [67]. Twenty-four hour dietary recalls have a series of inherent limitations: they may not reflect usual intake and they are memory dependent, which may lead to under-or over-reporting; however, a single 24-hour recall is sufficient to report mean group intake [68]. For children two to 11 years, proxies reported or assisted with the 24-hour recalls. Proxies, usually parents, can often report accurately what their children eat at home [69]; however, they may not know what they eat outside the home [70], which could result in reporting errors [71]. Use of standardized recipe files and food composition databases can also increase error when examining nutrient intake [72].
This study, as are all studies of breakfast, was limited by the lack of a standard definition of "breakfast" and of "breakfast skippers" [30]. In this study breakfast was self-defined; further, categorization of a child into a breakfast pattern or as a breakfast skipper was based on consumption the day of the recall only with may have led to misclassification of children into a particular breakfast pattern or as a breakfast skipper. The association of consumption of individual breakfast patterns on total daily intake may not reflect the breakfast patterns per se, but may be the result of what is consumed throughout the day. Lastly, there is a potential for residual confounding or to other variables not measured in the NHANES [73].

Conclusion
These data indicate that the breakfast meal has the potential to be an important meal and generally makes a positive contribution to nutrient intake, diet quality, and weight/adiposity parameters, but that care should be taken by individuals to select nutrient-dense foods, such as fruit/FJ, LFM, and fortified grain foods low in fat and added sugars. It is also important to integrate this type of nutrient-dense breakfast with an overall healthy eating plan [74]. These results, as well as the inverse association with weight and adiposity measures shown in consumers placed in some patterns, need further study to determine how breakfast meals influence energy, nutrient, and food group intakes and the timing of meals/snacks throughout the day. work reported or in the conclusions, implications, and opinions stated. All authors contributed equally to this manuscript. TAN, CO'N, and VLF planned the study; VLF conducted the statistical analyses; TAN, CO'N, and VL interpreted the data; CO'N wrote the original draft of the manuscript; TAN and VLF edited this manuscript.