European Journal of Clinical Nutrition (2010) 64, 184193

& 2010 Macmillan Publishers Limited All rights reserved 0954-3007/10 $32.00
www.nature.com/ejcn

ORIGINAL ARTICLE

The association of maternal diet and dietary supplement intake in pregnant New Zealand women with infant birthweight
PE Watson1 and BW McDonald2
1 Institute of Food Nutrition and Human Health, Massey University, Albany Campus, Auckland, New Zealand and 2Institute of Information and Mathematical Sciences, Massey University, Albany Campus, Auckland, New Zealand

Objective: To investigate the association of infant birthweight with maternal diet and supplement intake. Subjects/Methods: Prospective cohort study of 504 European and Polynesian urban and rural pregnant volunteers recruited from northern New Zealand clinics. Subjects were visited in months 4 (mth4) and 7 (mth7) of pregnancy when height, weight and skinfolds were measured, questionnaires to determine personal details administered, and diet assessed by a 24-hour recall and 3-day food record. Results: After adjusting for confounders nutrients accounted for up to 5.0% of the total variance in birthweight. Ethnicity was not a significant confounder. A quadratic relationship existed between birthweight and % total energy (%TE) from carbohydrate, fat and protein, most significantly with carbohydrate energy (P 0.002). Birthweight was greatest (B3600 g) when carbohydrate %TE was 48%, fat 35% and protein 17%. Birthweight was reduced with high b-carotene intakes (mth4, P 0.009) and with both high retinol and b-carotene intakes in mth4 and 7 (average). Birthweight was positively associated with increasing pantothenic acid/biotin ratios (P 0.011), magnesium (P 0.000) and vitamin D (P 0.015) intakes in mth4; with biotin (P 0.040) and B12 intakes above the RDI (P 0.006) in mth7; and with pantothenic acid intake in mth4&7 (P 0.002). Dietary supplement usage was associated with increased birthweight, most significantly iron supplementation (P 0.006). Conclusion: Birthweight was associated with the %TE from carbohydrate, fat and protein, and with b-carotene, retinol, vitamins D and B12, pantothenic acid, biotin and magnesium intakes and iron supplementation. More research may be required on some dietary recommendations for pregnancy.

European Journal of Clinical Nutrition (2010) 64, 184193; doi:10.1038/ejcn.2009.134; published online 18 November 2009
Keywords: pregnancy; nutrient intake; dietary supplements; birthweight

Introduction
There is little agreement between published dietary studies conducted during pregnancy on which nutrients in the maternal diet are associated with infant weight at birth after

Correspondence: Dr BW McDonald, Institute of Information and Mathematical Sciences, Massey University, Albany Campus, Private Bag 102904, North Shore Mail Centre 1311, Auckland, New Zealand. E-mail: B.McDonald@massey.ac.nz Guarantors: PE Watson and BW McDonald. Contributors: PEW designed the study, recruited the subjects, collected the data, analyzed the dietary data, and wrote the manuscript. BWM analyzed the data and wrote the manuscript. Received 21 January 2009; revised 22 September 2009; accepted 2 October 2009; published online 18 November 2009

adjustment for confounders. Haste et al. (1991) collected weighed diet records from 169 women and found no association between diet and birthweight at 28 weeks, but at 36 weeks found protein, zinc, thiamine and riboflavin intakes and the change in intake between 28 and 36 weeks of these nutrients and iron had a significant positive effect on birthweight explaining between 2.4% and 7.2% of the total variance. Godfrey et al. (1996) administered a food frequency questionnaire (FFQ) to 538 women in early and late pregnancy and found that increasing carbohydrate intake in early pregnancy and increasing meat protein intake in late pregnancy was associated with decreased birthweight. Mathews et al. (1999) studied 693 subjects who completed a food diary in early pregnancy and a FFQ after 28 weeks and found higher vitamin C intakes in early pregnancy were

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

185
associated with increased birthweight. Moore et al. (2004) administered a FFQ to 429 women in early and late pregnancy and found the percentage of energy derived from protein, particularly dairy protein in early pregnancy, was positively associated with birthweight. Other small studies have found no association between diet and birthweight (Langley-Evans and Langley-Evans, 2003; Lagiou et al., 2004). Doyle et al. (1990), in a study that did not correct for confounding variables, found that lower maternal intakes of B vitamins especially thiamine and niacin, and minerals especially magnesium and iron were significantly associated with lower birthweights. Several studies have used retrospective FFQ administered after birth to determine dietary intake during pregnancy. Petridou et al. (1998) found higher meat and fish intakes were suggestively associated, and MUFA intake significantly associated with increased birthweight. Mitchell et al. (2004) found that intake of iron supplements in late pregnancy was associated with increased birthweight whereas Andreasyan et al. (2007) found increasing protein intake was negatively associated with birthweight. The aim of our study was to investigate the association of maternal diet and supplement intake with infant birthweight in European and Polynesian women after adjusting for all possible confounders. Weight was measured on portable Tanita 1609 electronic scales, calibrated weekly against a Sauter platform scale accurate to 2000.02 kg. Height was measured with a custom made portable stadiometer. Triceps, biceps and costal skinfolds were measured using HarpendenHoltain calipers. All measurements were according to the procedures set out in Gibson (1990). Birthweight was taken from the infants clinical record. Gestational age was calculated from the last date of menstruation. Dietary intake was assessed in mth4 and mth7 by 24-h recall followed by a 3-day food record. In the recall, the interviewer used a number of aids to assess the size and volume of each food portion. The recalls included questions on the brand name and dose of any dietary supplements taken that day. For 3 days after each recall, subjects recorded in their preferred language all foods and drinks consumed using the provided measuring cups and spoons to assess volume ingested. Each 4 days of diet assessment included one weekend day. Analysis of variance of nutrient intake in both mth4 and mth7 diets found no significant difference overall between the 24-h recall and 3-day food records in each period (P 0.099, Pillais test), neither were there any significant differences overall in nutrient density between methods (MANOVA P 0.158; Watson and McDonald, 2003). Consequently the 24-h recall and 3-day diet record data were combined to provide the mean intake of each nutrient for each woman, in mth4 and mth7 of pregnancy. No correction was made for under-reporting because of the unpredictable changes in basal metabolic rate during pregnancy (Kopp-Hoolihan et al., 1999; Prentice and Goldberg, 2000) and the 73% of subjects suffering from morning sickness in early pregnancy. Foodworks (Xyris Software utilizing the New Zealand Food Composition database FOODfiles, New Zealand Institute for Crop & Food Research) was used to calculate the nutrient intake for each woman. The New Zealand Institute of Environmental Science and Research Dietary Supplement database, which provided manufacturers information on supplement composition and dosage, was used to calculate the nutrient load provided by supplements taken by each woman. Where the brand of supplement was not known the generic dose was used. Some supplements were made by reputable manufacturers with good quality control, but others were not. There was no independent chemical analysis to verify the stated nutrient dose, and losses during storage of labile nutrients may have occurred. Thus the nutrient content of supplements may have varied considerably from the stated dose (Larimore and Mathuna, 2003). In addition, bioavailability of supplement nutrients may have differed to nutrients from food. For these reasons supplement nutrients were treated as separate variables to food nutrients, which also means that their effects could be assessed separately. Maternal body measurements at mth4 and mth7 investigated as potential confounders were: height, age, weight, body mass index (BMI), sum of skinfolds and weight gain European Journal of Clinical Nutrition

Methods
Ethical approval was obtained from the Massey University Human Ethics Committee and the Auckland Ethics Committee. Subjects were drawn from urban and rural centers in the upper North Island of New Zealand. The funding body required 500 subjects with selection deliberately biased towards women of lower socioeconomic status, and including a greater proportion of Maori and Pacific women than in the general population. After widespread publicity, targeted clinics in lower socioeconomic areas throughout the study area distributed study information to potential volunteers, and 504 women around the 14th week of pregnancy with no record of chronic disease or history of obstetric problems were recruited. The results and tables in this paper are based on data from 439 subjects who had singleton births and for whom we had demographic, anthropometric, nutrient intake and birth outcome details. Data for two women with infants born more than 8 weeks premature were not used. Subjects were visited by an interviewer of their own ethnicity near the start of mth4 and mth7 of their pregnancy. Questionnaires were administered in the subjects preferred language to determine demographic, medical, well-being and lifestyle details. They were similar in content and format to questionnaires used to determine this information in national nutrition surveys in New Zealand (Quigley and Watts, 1997). The information used is described in the analysis section.

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

186
between mth4 and mth7. Weight was linearly adjusted to equate to that at 18 and 30 weeks of gestation and these were used to calculate weight gain between 18 and 30 weeks. Other maternal variables considered were the total of other adults in the household (for example partner, relatives, boarders), presence of a partner, ethnicity, parity, number of preschoolers (an indicator of the time between pregnancies), years of schooling, years of tertiary education, partners or subjects occupation, welfare status, income level, number of cigarettes smoked each day, morning sickness severity, and presence of hypertension, gestational diabetes and anaemia. Infant variables considered as potential confounders were gestational age of the newborn (including an adjustment if the baby was born more than 7 days late as birthweight tended to plateau after this time), and sex of the infant. Data were checked using standard statistical techniques (for example crosstabs, scatterplots). Minitab 15 was used to analyze the data. Results were regarded as significant at Po0.05. Nutrients were expressed on a logarithmic scale to reduce the impact of outliers and to investigate relative differences between low and moderate intakes. Use of untransformed nutrients did not improve the models. Standard model selection procedures (stepwise and best subsets regression) were followed to identify significant associations between nutrient data, confounders and birthweight, and all associations carefully checked to ensure they were not due to outliers or multicollinearity. First the confounders that remained significant in all exploratory stepwise regressions were retained in the analyses. Second birthweight was regressed against each nutrient singly, and the confounders with each nutrient singly: for mth4 dietary data (n 424), mth7 dietary data (n 403) and for a combination of mth4&7 data (n 439). The latter was to give an indication of maternal diet throughout pregnancy. In 88% of cases the combination was the average of mth4 and mth7 data or, in the few cases where one or the other was not available, the average data from the available month. Finally an overall regression model including the nutrients was selected for mth4, mth7 and mth4&7 data. A similar procedure was used to investigate the association of birthweight with dietary supplement intake, with and without including dietary nutrient intake in the model. birthweight associated with each confounder are shown in Table 2. Table 3 shows the mean nutrient intake from food in mth4 and mth7 and the significant differences by paired t-test between mth4 and mth7 intakes. There were fewer significant differences in nutrient density between mth4 and mth7. Table 3 also shows the significant associations of birthweight versus the basic confounders and log of each nutrient singly for mth4, mth7 and a combination of mth4&7. The greatest number of significant associations was found for the combined mth4&7 data (25), then the mth4 data (22), with few associations being found for mth7 data (3) and even fewer for supplements alone (1). Nutrient density models for birthweight were considered, by including log energy intake as a confounder in the calculations, but the associations of nutrient intakes with birthweight were invariably weaker, indicating that the nutrient effect was due to intake quantity rather than quality of diet. Regression analysis was then used to find the best model for birthweight in terms of the confounders and nutrients. The nutrients significantly associated with birthweight are shown in Table 4. If alcohol intake during mth4 (P 0.039) was included with the combined mth4&7 data the variance explained by diet increased from 3.7% to 4.4%. In all models (mth4, mth7 and combined mth4&7 diets), diet quality in terms of the % total energy (%TE) from carbohydrate, fat and protein was significantly associated with birthweight. Further analysis revealed this relationship was quadratic. This is illustrated in Figure 1 which shows the results of polynomial regression analyses of birthweight versus %TE from carbohydrate, fat, protein and protein intake in grams. Similar curvature was found after adjusting for confounders. The strongest expression of this association was between birthweight and %TE from carbohydrate: P 0.009 for the quadratic term before adjusting for confounders, and P 0.032 after adjusting for confounders. Average birthweight maximized around 3600 g when the %TE from carbohydrate was 48%, 35% from fat and 17% from protein. The quadratic trend for protein %TE was weaker (quadratic P 0.077), whereas the quadratic term for log(protein) was significant at P 0.043. In the multiple regression analyses all these relationships with birthweight could be accounted for by a single term for carbohydrate (%TE from carbohydrate48)2. Below and above 48%TE from carbohydrate, birthweight decreased. At the lower quartile for energy from carbohydrate (41%TE), mean fat intake contributed 42.5% and protein 16.3% of TE (mean birthweight 3518 g). When energy from carbohydrate was at the upper quartile (55.8%TE) mean fat intake contributed 29.9% and protein intake 14.5% of TE (mean birthweight 3504 g). At the extremes of TE from carbohydrate (31% and 71%), predicted birthweight was respectively 291 and 617 g less than the maximum of 3600 g. Either pantothenic acid, biotin or the pantothenic acid/ biotin ratio were significant and positively associated with increased birthweight. Figure 2 shows the association with

Results
Seventy-five percent of subjects classified themselves as being of European origin, 18% as Maori and 7% as being Pacific Polynesian. The demographic and anthropometric details for all subjects and their infants, and by ethnic group, are shown in Table 1. Many of the potential confounders did not remain significant in the stepwise regressions against birthweight, for example ethnicity was not a significant confounder once maternal weight and height were taken into account. The significant confounders and the gram per unit change in European Journal of Clinical Nutrition

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

187
Table 1 Demographic and anthropometric characteristics of all subjects and infants and by ethnic group All subjects (N 439) Demographic characteristics Household size (mean, s.d.) Subjects married/partnered (%)d Children in household (mean, s.d.) High-occupation group (%)h Low-occupation group (%)h Welfare group (%)h Years secondary schooling (mean, s.d.) Some tertiary education (%) Subject in paid work (%) Subjects smoked in pregnancy (%) Maternal measures (mean, s.d.) Age (years) Height (cm) Weight before pregnancy (kg)e Weight mth4 (kg) Weight mth7 (kg) BMI mth4 (kg/m2) BMI mth7 (kg/m2) Sum skinfolds mth4 (mm) Sum skinfolds mth7 (mm) Infant measures (mean, s.d.) Birthweight (g) Head circumference (cm)f Length (cm)f European (N 329) Maori (N 80) Pacific (N 30) P-value

3.4 (1.4) 92.9 1.1 (1.2) 40.9 50.2 8.9 4.1 (0.9) 78.2 64.1 11.4

3.1 (1.1) 97.3 0.9 (1.0) 48.6 46.5 4.9 4.2 (0.9) 80.6 69.6 5.5

4.1 (1.5) 77.8 1.6 (1.4) 19.7 58.0 22.2 3.9 (1.1) 75.3 46.9 35.8

4.8 (2.0) 86.7 1.8 (1.7) 13.3 70.0 16.7 4.0 (1.0) 60.0 50.0 10.0

0.000a 0.000b 0.000a 0.000b 0.000b 0.000b 0.050a 0.026b 0.000b 0.000b 0.000c 0.009c 0.000c 0.000c 0.000c 0.000c 0.006c 0.000c 0.000c 0.026c 0.553c 0.515c

31.2 167.5 66.4 71.8 77.7 26.2 29.9 54.6 55.8

(5.1) (6.1) (13.5) (13.7) (13.4) (4.5) (4.7) (19.5) (18.4)

32.1 167.1 65.2 69.9 76.1 25.6 29.6 51.2 52.7

(4.4) (6.0) (13.0) (12.1) (12.1) (4.1) (4.30) (16.6) (15.8)

28.2 167.9 68.7 74.4 83.4 27.3 30.9 62.2 65.5

(6.4) (6.6) (13.1) (15.3) (15.8) (4.8) (5.7) (24.4) (22.5)

29.0 170.6 78.0 85.9 88.5 30.4 33.0 75.5 73.8

(5.6) (4.9) (15.4) (16.7) (18.5)g (5.3) (6.6)g (19.7) (21.6)g

3551 (544) 35.1 (1.6) 52.0 (2.7)

3551 (532) 35.1 (1.5) 52.0 (2.7)

3467 (581) 35.1 (2.0) 52.1 (2.7)

3780 (528) 35.4 (1.6) 52.8 (2.4)

Significant differences between ethnic groups are shown. a KruskalWallis test. b 2 w test for homogeneity. c Analysis of variance. d Almost all of the partners worked full time. e As recalled by subject, not measured. f Head circumference and length were not recorded for all infants. g Sample size reduced to 15, as 15 of the heavier women did not consent to being measured. h The upper-occupation group includes categories 13, and the lower-occupation group categories 410 of the ISCO 88 (International Standard Classification of Occupation) for each subjects partner. In few cases where the partner was unemployed but the subject was employed, occupational category of the subject was used. Welfare group included unemployed men with unemployed partners, or unemployed solo women on welfare.

Table 2 Regression coefficients for infant birthweight versus significant confounders Significant confounders Gram per unit change in birthweight (s.e.) (N 439) 25 (2) 15 (6) 135 (42) 9 (3.7) 7.3 (1.6) 98 (30) 97 (30) 71 (29) 35.8% P-value

Gestational age (days) Reduction in age slope if 47 days late (days) Male infanta Maternal height (cm) Maternal mth4 weight (kg) Smoking categoryb Number preschoolers Number other adults in house R2
a b

0.000 0.022 0.001 0.022 0.000 0.001 0.001 0.015

0 no, 1 yes. 0 nonsmoker, 1 15 cigarettes per day, 2 610 cigarettes per day, 3 11 cigarettes per day.

pantothenic acid. In the averaged mth4&7 data the interquartile effect of pantothenic (difference in birthweight between women on the third versus the first quartile of

pantothenic) was 83 g. In mth4 the pantothenic acid/biotin ratio was important (interquartile effect 62 g) and if biotin intakes were high it was beneficial to birthweight to also have a high pantothenic intake. This effect never approached significance in mth7. Biotin was significant in mth7 (interquartile effect 58 g) and if pantothenic was included in the model instead of biotin its effect on birthweight was positive but not significant. In mth4 increased b-carotene intakes were associated with decreased birthweight (interquartile effect 76 g), but there was no association in mth7. For the combined mth4&7 data, b-carotene and retinol together had a significant effect on birthweight, succinctly represented by dividing the data into four groups in terms of low b-carotene (first quartile, p2000 mg) versus the rest, and low retinol (first quartile, p315 mg) versus the rest. This four-group variable was significant (P 0.032) and its effect on birthweight is illustrated in Figure 3. In mth4 magnesium was positively associated with birthweight (Figure 2), interquartile effect, 122 g. The European Journal of Clinical Nutrition

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

188
Table 3 Mean mth4 and mth7 daily nutrient intakes and the significant associations of birthweight versus the confounders and log of each nutrient singly for mth4, mth7 and the mean of log mth4&7 intakes Nutrient Mean mth4 nutrient intake (N 424) Mean mth7 nutrient intake (N 403) Birthweight versus confoundersa and mth4 log nutrient data (N 424) P-value Birthweight versus confoundersa and mth7 log nutrient data (N 403) P-value Birthweight versus confoundersa and mean mth4&7 log nutrient data (N 439) P-value 0.045 0.026 0.003

Energy (kJ) Alcohol (g) Protein (g) Carbohydrate (g) Glucose (g) Fructose (g) Sucrose (g) Lactose(g) Maltose (g) Starch (g) Fibre (g) Cholesterol (mg) Fat (g) SFA (g) MUFA (g) PUFA (g) b-Carotene (mg) Retinol (mg) Vitamin A (mg) Vitamin C (mg) Vitamin D (mg) Vitamin E (mg) Thiamin (mg) Riboflavin (mg) Niacin equivalent (mg) Vitamin B6 (mg) Vitamin B12 (mg) Pantothenic (mg) Biotin (mg) Folate (mg) Sodium (mg) Potassium (mg) Magnesium (mg) Calcium (mg) Phosphorus (mg) Manganese (mg) Iron (mg) Zinc (mg) Sulphur (mg) Chloride (mg) Copper (mg) Selenium (mg)

9461 2.4 86* 270 26 27 61 17 3 136* 24* 306** 92 41 29 12 3672 462 1075 154 2.1 9.9* 2.0 1.9 17.7* 1.9* 4.6 5.3 40.4 285 3193** 3506 328 936 1485 4806 12.6 11.6** 933** 4944** 1.6 67

9204 2.5 83* 267 26 27 63 18 3 129* 23* 271** 88 40 28 11 3913 485 1138 153 2.1 9.5* 2.3 1.9 16.7* 1.8* 4.2 5.2 38.9 276 2987** 3448 317 953 1450 4600 12.2 10.8** 885** 4626** 1.5 64

0.031 0.031 0.018

0.012 0.021 0.006 0.029

0.045

0.007 0.043 0.005 0.018 0.016 0.002 0.037 0.024 0.017 0.003 0.005 0.002 0.009 0.016 0.029 0.011 0.006

0.029 0.025 0.003 0.008 0.009 0.001 0.013 0.022 0.015 0.001 0.001 0.009 0.001 0.012 0.006 0.010 0.008 0.004 0.002

0.034 0.019

0.032

Abbreviations: MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; SFA, saturated fatty acids. *Mean intake significantly different at Po0.05. **Significantly different at Po0.01. a Confounders: gestational age (modified for overterm babies), infant gender, maternal height, weight, smoking, number of preschoolers, number of other adults in house.

interquartile effect of vitamin D dietary intake on birthweight was 71 g but intakes were very low (median 1.7 mg) and we could not estimate the amount synthesized in the skin. The negative associations of alcohol and sucrose intakes with birthweight were only significant after adjustment for other nutrients, which makes estimation of their effects problematic. Selenium intakes were also associated with decreased birthweight, but further analysis showed that the negative trend was due to a few people with very high intakes (4300 mg). European Journal of Clinical Nutrition

In mth7 increased intakes of vitamin B12 had a positive association with birthweight across the full range of intakes (P 0.009) but the association was stronger above the RDI of 2.6 mg (P 0.006) as illustrated in Figure 4. Biotin was only significant in the presence of the confounders. Cholesterol was only significant if the confounders and B12 or biotin were in the model. Up to three different dietary supplements were taken by 31.5% of women (mostly Europeans) in mth4, and 31.3% in mth7. The most common were iron, multivitamins and

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

189
Table 4 Significant associations birthweight versus mean daily nutrient intake after adjustment for confounders: at mth 4, mth7 and mth4&7 Significant nutrients after confoundersa Mean mth4 and 7 diets (N 439) Signed b P-value 0.002 0.002b 0.039 Mean mth4 diets (N 424) Signed P-value Mean mth7 diets (N 403) Signed P-value

Log pantothenic acid Sq(carb/nrg%48) b-Carotene retinol groups Log b-carotene Log (pantothenate/biotin) Log biotin Log Mg Log(1 alcohol mth4) Log sucrose Log vitamin B12 Log cholesterol Log vitamin D Log Se R2 nutrients only
a

0.026 0.009 0.011

0.053

0.040 0.000 0.024 0.032 0.009 0.003 0.015 0.029 5.0 %

3.7%

3.1%

Confounders: gestational age (modified for overterm babies), infant gender, maternal height, weight, smoking, number of preschoolers, number of other adults in house. b Negative sign attached to P-value indicates a decreasing association.

Birth Weight = 1176 + 102 Carb%Energy - 1.07 Carb%Energy**2

Birth Weight = 2297 + 70.6 Fat%Energy - 0.97 Fat%Energy**2

5000 Birth Weight (g) Birth Weight (g) 30 40 50 60 70 4000 3000 2000 1000 Carbohydrate % of Energy

5000 4000 3000 2000 1000 20 30 40 50 60

Fat % of Energy

Birth Weight = 2423 + 136 Protein%Energy - 3.96 Protein%Energy**2 5000

Birth Weight = 2926 + 12.3 Protein - 0.053 Protein**2

5000 Birth Weight (g) 4000 3000 2000 1000 10 15 20 25 0 50 100 Protein (g) 150 200

Birth Weight (g)

4000 3000 2000 1000

Protein % of Energy

Figure 1 Polynomial regression analysis: birthweight versus (a) carbohydrate % total energy (%TE), (b) fat %TE, (c) protein %TE and (d) protein (g) in mean mth4&7 data.

European Journal of Clinical Nutrition

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

190

Baby Weight vs Pantothenate BabyWeight = 3190 + 524.4 log10(Pantothenate) R-sq 2.1% 6000

Baby Weight vs Magnesium BabyWeight = 2291 + 508.7 log10(Magnesium) R-sq 1.7% 6000

5000 Baby Weight g

5000

BabyWeight g

4000

4000

3000

3000

2000 RDI 1 5 Pantothenate mg

Figure 2

2000 RDI 10 100 200 Magnesium mg 500

Birthweight versus (a) pantothenic acid intake in mth4&7 and (b) magnesium intake in mth4.

Effect of B-Carotene and Retinol on Birthweight After adjusting for carbohydrate, pantothenic and all confounders 3660 3640 Mean Birthweight g 3620 3600 3580 3560 3540 3520 3500 0 Retinol >315 ug 1 high BC, high R low BC, low R Birthweight (g) high BC, low R low BC, High R
B-Carotene >2000 ug

Scatterplot of Birthweight vs log10(vitamin B12 intake mth7) 0.415 5500 5000 4500 4000 3500 3000 2500 2000 -0.5 0.0 0.5 log10(vitamin B12) 1.0 1.5 0 1 Below/AboveRDI 0 1

Figure 3 Effect of mean mth4&7 intake of b-carotene and retinol on birthweight, after adjusting for carbohydrate, pantothenic and all confounders. Note: low b-carotene, low retinol group mean vitamin A intake was 366 mg (n 28); low b-carotene, high retinol group mean vitamin A intake was 986 mg (n 78); high b-carotene, low retinol group mean vitamin A intake was 817 mg (n 73); high b-carotene, high retinol group mean vitamin A intake was 1289 mg (n 261). Confounders: gestational age (modified for overterm babies), infant gender, maternal height, weight, smoking, number of preschoolers, number of other adults in house.

Figure 4 Birthweight versus mean vitamin B12 intake below and above the RDI (2.6 mg) in mth7.

supplement nutrient that had a significant association with birthweight ( 119 g, P 0.006).

minerals, and folate. It was more significant that a supplement was taken, than the actual dose. After adjusting for confounders and base nutrients there was a 129-g increase in birthweight (P 0.002) in those who took dietary supplements compared to those who did not. Iron was the only European Journal of Clinical Nutrition

Discussion
This study has shown for the first time that the relationship between birthweight and energy sources in the diet is not linear but quadratic with a maximum corresponding to

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

191
48%TE from carbohydrate, 35% from fat and 17% from protein. In terms of the New Zealand and Australian RDI for pregnant women these figures are at the lower end of the recommended range of 4565%TE from carbohydrate and 1525%TE from protein and at the top of the recommended range of 2035%TE from fat (NHMRC, 2006). Mean birthweight was at its lowest (2971 g) when carbohydrate energy was at its highest (71%). It is possible that diets with a high proportion of energy from carbohydrate may overload the adenosine triphosphate system, increase free radical production, increase DNA damage and thus decrease cell division and growth. Energy, carbohydrate and fat intakes were highly correlated, with protein being more distant. Studies in developed countries have found an association between birthweight and intake of a single macronutrient or the %TE from a macronutrient. Godfrey et al. (1996) found increasing carbohydrate intake in early pregnancy was associated with a decrease in birthweight. A positive association with birthweight was found for %TE from carbohydrate in early pregnancy (Cohen et al., 2001), monounsaturated fat intake in late pregnancy (Petridou et al., 1998), protein intake in early pregnancy (Doyle et al., 1990; Cuco et al., 2006), %TE from protein in early pregnancy (Moore et al., 2004) and protein intake in late pregnancy (Haste et al., 1991; Cuco et al., 2006). Other studies have found birthweight declines with high protein intakes (Andreasyan et al., 2007). These varying results are not inconsistent considering the close correlation of these macronutrients and the quadratic relationship between carbohydrate, protein and fat intakes and birthweight found in our study. Sloan et al. (2001) reported a quadratic relationship with protein alone. Their quadratic showed a weak effect of low protein whereas our data suggests a much stronger deleterious association of low protein intake with birthweight. The quadratic curve maximum was at 70 g in Sloans sample and 108 g in ours (Figure 1), both well above the 60 g RDI for protein for pregnant women (NHMRC, 2006). Pantothenic, biotin and magnesium intakes are strongly correlated (r 0.750.80). Increasing intakes in all were associated with increased birthweight. These nutrients all have essential roles in cellular energy metabolism. When compared to their respective RDIs, 25% of subjects consumed less than 30 mg of biotin, 48% consumed less than 5 mg of pantothenic and 69% of subjects consumed less than 360 mg of magnesium. Of these magnesium intakes had the greatest association with birthweight and biotin the least. Low maternal magnesium intakes were found to be associated with lower birthweight in a study where no correction was made for confounders (Doyle et al., 1989). Merialdi et al. (2003) reported that magnesium supplementation during pregnancy resulted in an increase in birthweight. Takaya et al. (2004) found that intracellular [Mg2 ] in cord platelets was significantly associated with birthweight and postulated that decreased fetal intracellular [Mg2 ] might underlie later development of insulin resistance. Figure 3 gives the mean vitamin A intake and number of subjects in each b-carotene/retinol grouping. Though the mean vitamin A intake in the low b-carotene/low retinol group was below the EAR of 550 mg this was not associated with a decrease in birthweight (NHMRC). Birthweight in all groups except those with both moderate-high b-carotene and retinol was above average (3529 g). The mean vitamin A intake in the latter group was only 489 mg above the RDI (800 mg) and well under the recommended UL of 3000 mg, yet there was a negative effect on estimated birthweight, which was below average and at least 80 g less than the other groups. Known fetopathic intakes are much higher, but perhaps the fetus is very sensitive to b-carotene/retinol levels and at these intakes prooxidant action may occur at a sufficient level to decrease birthweight. This may also explain the negative effect of increasing b-carotene intakes in mth4 where some intakes were very high. To date a prooxidant effect has only been reported in in-vivo studies when high levels of dietary supplements are taken or shown in in-vitro studies of cell function (Omenn et al., 1996; Palozza et al., 2003). Mathews et al. (2004) found high serum retinol levels in late pregnancy were associated with lower birthweight (Po0.001). Vitamin D synthesis in the skin may have been inadequate in many subjects in this study as dietary vitamin D intake was positively associated with birthweight. Few foods are supplemented with vitamin D in New Zealand. Mannion et al. (2006) found that intake of milk fortified with vitamin D, and intake of vitamin D were each positively associated with birthweight. Olsen et al. (2007) also found that milk and milk protein intake was associated with higher birthweight and suggested that water soluble substances in milk increase fetal growth. They did not investigate vitamin D intake. Hollis and Wagner (2006) suggest the scientific basis for the current RDI for vitamin D is inadequate and that its actions as a potent steroid hormone need to be taken into account. Newly absorbed dietary B12 is preferentially transported through the placenta to the fetus, rather than stored vitamin B12 from the liver. If dietary intake is insufficient transfer to the fetus may be inadequate (Relton et al., 2005). Low maternal serum B12 is associated with low birthweight in India (Muthayya et al., 2006). In our study intakes of B12 were high with only 16% of women with vitamin B12 intakes oRDI (2.6 mg) and 9% with intakes oAI (2.2 mg), but birthweight only increased with intakes well above the RDI (Figure 4). In terms of foods, these results suggest pregnant women should choose diets with moderate amounts of breads and cereals preferably whole grain, and low intakes of refined sweet foods for example confectionary, sweet drinks, cakes and biscuits. Diets should also contain moderate quantities of protein foods for example lean meat, lean poultry, eggs, fatty fish and low fat milk products. Milk products and margarines fortified with vitamin D are preferred. Generous intakes of green vegetables, legumes and nuts are desirable. Alcoholic beverages should be avoided. European Journal of Clinical Nutrition

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

192
In conclusion, significant associations were found between maternal diet and infant birthweight. Nutrients accounted for 5.0% of total variance in birthweight in mth4, 3.7% in the combined mth4&7 data, and 3% in mth7. A new finding was the quadratic relationship between birthweight and %TE from carbohydrate, fat and protein throughout pregnancy and that birthweight was greatest when carbohydrate %TE was 48%, fat 35% and protein 17%. High carbohydrate energy diets had a greater negative effect on birthweight than low. Increased birthweight was associated with increasing pantothenic, biotin, magnesium, vitamin D and vitamin B12 intakes. High b-carotene and retinol intakes were associated with decreased birthweight. Dietary supplement intake, especially iron supplementation, was associated with an increase in birthweight. These results suggest that more research on diet during pregnancy is required to: confirm the RDIs for %TE from carbohydrate, fat and protein; determine whether an UL is required for b-carotene; investigate whether the UL for retinol intake is set too high, and whether the RDI for B12 is too low.
Godfrey K, Robinson S, Barker DJP, Osmond C, Cox V (1996). Maternal nutrition in early and late pregnancy in relation to placental and fetal growth. BMJ 312, 410414. Haste FM, Brooke OG, Anderson HR, Bland JM (1991). The effect of nutritional intake on outcome of pregnancy in smokers and nonsmokers. Brit J Nutr 65, 347354. Hollis BW, Wagner CL (2006). Nutritional vitamin D status during pregnancy: reasons for concern. CMAJ 174, 12871290. Kopp-Hoolihan LE, van Loan MD, Wong WW, King JC (1999). Longitudinal assessment of energy balance in well-nourished, pregnant women. Am J Clin Nutr 69, 697704. Lagiou P, Tamimi RM, Mucci LA, Adami H-O, Hsieh C-C, Trichopoulos D (2004). Diet during pregnancy in relation to maternal weight gain and birth size. Eur J Clin Nutr 58, 231237. Langley-Evans AJ, Langley-Evans SC (2003). Relationship between maternal nutrient intakes in early and late pregnancy and infants weight and proportions at birth: prospective cohort study. J R Soc Promotion Health 123, 210216. Larimore WL, Mathuna DP (2003). Quality assessment programs for dietary supplements. Ann Pharmacother 37, 893898. Mannion CA, Gray-Donald K, Koski KG (2006). Association of low intake of milk and vitamin D during pregnancy with decreased birthweight. CMAJ 174, 15. Mathews F, Yudkin P, Neil A (1999). Influence of maternal nutrition on outcome of pregnancy: prospective cohort study. BMJ 319, 339343. Mathews F, Youngman L, Neil A (2004). Maternal circulating nutrient concentrations in pregnancy: implications for birth and placental weights in term infants. Am J Clin Nutr 79, 79103. Merialdi M, Carroli G, Villar J, Abalos E, Gulmezoglu AM, Kulier R et al. (2003). Nutritional interventions during pregnancy for the prevention or treatment of impaired fetal growth: an overview of randomized controlled trials. J Nutr 133, 1626S1631S. Mitchell EA, Robinson E, Clark PM, Becroft DMO, Glavish N, Pattison NS et al. (2004). Maternal nutritional risk factors for small for gestational age babies in a developed country: a case-control study. Arch Dis Childhood Fetal Neonatal Ed 89, F431F435. Moore VM, Davies MJ, Willson KJ, Worsley A, Robinson JS (2004). Dietary composition of pregnant women is related to size of the baby at birth. J Nutr 134, 18201826. Muthayya S, Kurpad AV, Duggan CP, Bosch RJ, Dwarkanath P, Mhaskar A et al. (2006). Low maternal vitamin B12 status is associated with intrauterine growth retardation in urban South Indians. Eur J Clin Nutr 60, 791801. NHMRC (2006). Nutrient Reference Values for Australia and New Zealand Including Recommended Dietary Intakes. Australia: National Health and Medical Research Council. Olsen SF, Halldorsson TI, Willett WC, Knudsen VK, Gillman MW, Mikkelson TB et al., NUTRIX Consortium (2007). Milk consumption during pregnancy is associated with increased infant size at birth: prospective cohort study. Am J Clin Nutr 86, 11041110. Omenn GS, Goodman GE, Thornquist MD, Malmes J, Cullen MR, Glass A et al. (1996). Effects of a combination of B-carotene and vitamin A on lung cancer and cardiovascular disease. NEJM 334, 11501155. Palozza P, Serini S, Di Nicuolo F, Piccioni E, Calviello G (2003). Prooxidant effects of b-carotene in cultured cells. Mol Aspects Med 24, 353362. Petridou E, Stoikidou M, Diamantopoulou M, Mera E, Dessypris N, Trichopoulos D (1998). Diet during pregnancy in relation to birthweight in healthy singletons. Child Care Health Dev 24, 229242.

Conflicts of interest
The authors declare no conflict of interest.

Acknowledgements
The authors wish to thank the New Zealand Ministry of Health and Health Research Council for their financial support, Mrs Chris King the project manager, the data collectors, data entry and analysis teams, especially Mrs Judi Scheffer, and most of all the subjects who gave willingly of her time.

References
Andreasyan K, Ponsonby AL, Dwyer T, Morley R, Riley M, Dear K et al. (2007). Higher maternal dietary protein intake in late pregnancy is associated with lower infant ponderal index at birth. Eur J Clin Nutr 61, 498508. Cohen GR, Curet LB, Levine RJ, Ewell MG, Morris CD, Catalano PM et al. (2001). Ethnicity, nutrition and birth outcomes in nulliparous women. Am J Obstet Gynecol 185, 660667. Cuco G, Arija V, Iranzo R, Vila J, Prieto MT, Fernandez-Ballart J (2006). Association of maternal protein intake before conception and throughout pregnancy with birthweight. Acta Obstet Gynecol Scand 85, 413421. Doyle W, Crawford MA, Wynn AH, Wynn SW (1989). Maternal magnesium intake and pregnancy outcome. Magnesium Res 2, 205210. Doyle W, Crawford MA, Wynn AH, Wynn SW (1990). The association between maternal diet and birth dimensions. J Nutr Med 1, 917. Gibson RS (1990). Principles of Nutritional Assessment. Oxford: Oxford University Press. pp 155260.

European Journal of Clinical Nutrition

Maternal Diet and Infant Birthweight PE Watson and BW McDonald

193
Prentice AM, Goldberg GR (2000). Energy adaptations in human pregnancy: limits and long term consequences. Am J Clin Nutr 71, 1226S1232S. Quigley R, Watts C (1997). Food Comes First: Methodologies for the National Nutrition Survey of New Zealand. Public Health Report Number 2, Ministry of Health: Wellington, New Zealand. Relton CL, Pearce MS, Parker L (2005). The influence of erythrocyte folate and serum vitamin B12 status on birthweight. Brit J Nutr 93, 593599. Sloan NL, Lederman SA, Leighton J, Himes JH, Rush D (2001). The effect of prenatal dietary protein intake on birthweight. Nutr Res 21, 129139. Takaya J, Yamata F, Higashino H, Kobayashi Y (2004). Relationship of intracellular magnesium of cord blood platelets to birthweight. Metabolism 53, 15441547. Watson P, McDonald B (2003). Nutrition During Pregnancy. Report to the Ministry of Health, Institute of Food Nutrition and Human Health, Albany Campus, Massey University, Auckland, New Zealand.

European Journal of Clinical Nutrition