Ear& Human Development, 10 (1984) 127-136 Elsevier

127

EHD 00588

The Bacon Chow study: effects of maternal nutritional supplementation on birth measurements

of children, accounting for the size of a previous (unsupplemented) child

William H. Mueller and Ernest0 Poll& School of Public Health, The University of Texas, Houston, TX 77025, U.S.A

Accepted for publication 6 April 1984

Summary

Studies on nutritional supplementation of pregnant mothers have observed small but sometimes significant effects on fetal growth. Previously we also reported significant changes in physical resemblance of siblings at birth due to this kind of supplementation, using data collected by the late Dr. Bacon Chow and associates. This suggested that components of variation in physical growth may change as a result of supplementation. To explore this further, we sought the effects of nutri- tional supplementation on somatic growth, holding constant the birth measurements of a previous (unsupplemented) sibling. Only head circumference at birth came close to showing a statistically significant (P < 0.06) supplement effect on its mean ( + 0.36 cm) when the measurement of the first infant was held constant.

Both increases and decreases in weight for length as assessed by Rohrer’s Index (wt/13) occurred with supplementation. Increases in the index occurred by a significant (P < 0.01) gain in birth weight (238 g) between 1st and 2nd siblings in the supplement group but not in the placebo group. Decreases in this index resulted from a significant gain in birth length (1.3 cm) in supplemented families with a previous (untreated) sibling having a below average length for average weight. These non-linear effects on somatic growth suggest that responses to maternal nutritional supplementation may be more effective in a subgroup of the population where the normal adaptive responses to low energy protein intakes may have failed or where nutritional conditions over pregnancy may have varied considerably.

nutrition; genetics; growth; Rohrer’s Index; anthropometry

Address all correspondence to: William H. Mueller, School of Public Health, The University of Texas, P.O. Box 20186, Houston, TX 77025, U.S.A.

0378-3782/84/$03.00 0 1984 Elsevier Science Publishers B.V.

128

Introduction

Effects of pre- and postnatal nutritional supplementation on mothers and children within populations nutritionally at risk have been assessed in various field studies [6,8,10,12]. Analyses to determine the effects of the supplementation have focused on comparisons of measures of central tendencies (i.e., means) of anthropometric measures between groups receiving different dietary treatment. Study designs in- cluded groups comprised of supplemented and unsupplemented subjects [10,12], or subjects that took different amounts of the same or similar supplements [6]. In one case [1,8], the design included subjects both supplemented and unsupplemented, as well as subjects that took different quantities of similar supplements.

The study referred to in this paper was initiated by the late Dr. Bacon Chow [1,3,8,11] of the Johns Hopkins School of Hygiene in 1967. This study was unique in that each participating mother provided two infants genetically related as full siblings. After the birth of the first infant, half the mothers received a high calorie (800 kcal) supplement with 40 g of protein daily (group A), while the other received a placebo supplement (group B). Treatment continued during the lactation of the first infant, through the gestation and lactation of a second infant. The first infant thus represents an untreated control.

A first report on the data from the Bacon Chow study showed that there were no differences between the mean birthweight of the supplement (group A; mean = 3118 g) and the placebo (group B; mean = 3073 g) infants [8]. However, within the A group, in those cases where the supplemented infant was a male, there were statistically significant differences between the mean birthweights of the supple- mented (mean = 3215 g) and the unsupplemented (mean = 3053 g) infants. When the supplemented infant was a female, there were no differences between first (mean = 3071 g) and second (supplemented - mean = 3012 g) offspring. Within the placebo group, none of the comparisons yielded statistically significant differences.

The correlations for the anthropometric measures at birth between the first and second offspring within the A and B groups have also been reported [ll]. In the placebo (B) group the correlations for all anthropometric measures, except skinfold thickness, were statistically significant, and at a level expected in most populations where there has been no selective mating (r = 0.50). However, within the A group the size of the first offspring was generally nof predictive of the second offspring, if this second offspring was a male. When the second offspring was a female, the correlations, except for Rohrer’s Index (wt/13), behaved very much as in the case of the placebo group (r = 0.50).

Comparisons of the results in the correlational analysis [ll] with those derived from the between-group comparisons [8] indicate that the correlations between offspring were more sensitive in detecting the effects of the supplement on growth than measures of central tendency. The sibling similarities within the supplemented (A) group, when the second offspring was a female, and that found in the placebo group, regardless of sex, were also taken as additional evidence that sex of the offspring moderates in part the effects of nutritional supplementation during preg- nancy.

129

If the effects of supplementation were similar across subjects then the sibling correlations in the high calorie group should have been the same as those observed in the placebo group. This raises the issue, already addressed cogently by others [7], that the effects of supplementation may not be linear and that there may be responders and non-responders to nutrition supplementation. Differential response to nutritional supplementation may be determined, among other factors, by non- nutritional factors that regulate in part the size of the fetus. This includes genetic makeup, maternal physiology and physique. In this paper we address ourselves in part to this issue by assessing the effects of supplementation on the birth measure- ments of a child, holding constant the measurement of his or her older sibling whose gestation was not supplemented.

Methods

Study design This study is a randomized, controlled double blind trial of nutritional supple-

mentation of pregnant and lactating women. 294 women randomly assigned to one of two treatment groups (A and B) represent the total sample. All women received a 12: oz. can of liquid supplement twice a day. The daily supplement for the A group provided 800 kcal and 40 g of protein. Group B (the controls) received a liquid which resembled the A group supplement in taste, texture and weight. Each can contained a total of 6 kcal. However in June 1971,4 years after the study began, the artificial sweetener in the placebo was replaced with sucrose, increasing the caloric content to about 40 kcal per can. Moreover, vitamins and minerals were also added in the last year of the study to the placebo. Only 24% of the B group second infants were born after the change in the treatment composition of the B supplement. The nutrient contents of the supplement and placebo are presented in Tables I and II. A tablet containing minerals and vitamins was delivered to the women in both groups. Treatment began 3 weeks after delivery of a first infant. This infant is the first (older) member of the sibling pairs dealt with in this paper. Supplementation continued during the lactation of this infant and the gestation and lactation of a

TABLE I

Nutrient content of one 12; oz. can of the supplement for groups A and B

Group A Group B a

Before 6/71 After 6/71

Protein 20 g * *

Fat 13.3 g * *

Carbohydrates 50 g * 10 g

Calories 400 kcal 3 kcal 43 kcal

a 76% of group B second study infants were born before 6/71, 52% were tested before 6/71. * Trace amount.

130

TABLE II

Vitamin and mineral content in one 12; oz. can of supplement A

Vitamin A 2500 USP units Vitamin D Vitamin C Thiamine Riboflavin Niacinamide Calcium Phosphorus Iron Vitamin E Pyridoxine Vitamin B12

200 USP units 31.5 mg 0.8 mg 0.9 mg 10.0 mg 0.5 g 0.4 g 6.0 mg 5.0 IU 0.8 mg

1.0 fig Calcium pantothenate Sodium Potassium Copper Manganese Fiber

4.0 mg 0.2 g 0.9 g 0.5 mg 1.0 mg 0.55 g

second infant - the younger member of the sibling pair. The treatment was not given to the infants directly.

The women for this study were recruited in 14 villages in Sui-Lin township, a farming area about 180 miles from Taipei, Taiwan. The 294 women selected for the study were judged to be in the lowest ranks of the socio-economic order. The average caloric intake was estimated to range from 1600 to 2000 kcal with protein intake of less than 40 g per day [5]. The staple diet consisted primarily of sweet potatoes and rice with little animal protein. Selection criteria also included that each woman be between 19 and 30 years of age, have at least one normal male child, be in the second or third trimester of pregnancy, and plan to have at least one more child.

Of the 294 women recruited, 225 gave birth to two infants within the 63 years spanned by the study. The 69 who failed to meet this minimum requirement were evenly divided between groups A and B. The present analysis was restricted to those cases with anthropometric measurements on both infants within 48 h of birth. Accordingly, 6 cases in each of the two groups were eliminated. The sample thus consists of 108 sibling pairs in the A group and 105 in the B group.

Data collection and analysis The data were collected in the field by teams of resident nurses who visited the

families periodically throughout the study. Nurses delivered the supplement twice daily. Compliance in consuming the supplement was monitored by measuring the amount of supplement remaining in the can after each delivery. Neither the nurses nor the women knew which supplement was delivered to a family.

The Sui-Lin study included the following birth measurements: weight; length; head, chest and abdomen circumferences; and triceps and subscapular skinfolds. We

131

considered all measurements except abdomen circumference and triceps skinfold. We found these latter measurements highly correlated with chest circumference and subscapular skinfold, respectively. An index of relative weight, Rohrer’s Index (wt/13), was also computed following the suggestion by Brandt [2] that this index best reflects intrauterine nutrition. It is not certain how many observers were involved in the anthropometry. However, the anthropometric records were checked daily for obviously erroneous values (Hsueh, personal communication).

Results

In Table III are presented the regression equations for the regression of birth measurements of a treated (supplement and placebo) child on those of his or her older sibling. The intercepts and slopes are shown with their standard errors to allow assessment of the significance of inter-group differences (supplement versus placebo). Because the effects of supplement on the sibling covariation in weight and head circumference appeared to be sex-dependent, regression equations are shown sep- arately by sex of the supplemented (younger) child in Table I. Among male children, differences between groups in the slopes and intercepts for weight, head cir- cumference and subscapular skinfold approach statistical significance (P = 0.11). For Rohrer’s Index of relative weight, group differences are statistically significant (P = 0.01). Among female children, the only significant difference between A and B groups is for Rohrer’s Index (P = 0.01). In all these cases, the slope of the placebo group is steeper and the intercept smaller than those of the protein-calorie supple- ment group.

A normal procedure to account for sibling covariance, would be to control for the variability due to a previous sibling and then test statistically for between-supple- ment-group differences in the means. However, in this case, the regression slope differs as a consequence of the supplementation. Accordingly, the proper adjustment equation is that of the placebo treatment [7]. Therefore, the B group equations were used to adjust the measurements of the supplemented child for those of a previous sibling and a one-way analysis of variance was done to test for between-group mean differences.

Among 12 comparisons (6 measurements x two sexes), there was only one mean difference which approached statistical significance (P < 0.06). The head cir- cumference of group A male children (protein-calorie supplement) was on the average 0.36 cm larger than placebo male children. Supplementation, thus, appears to affect more the variation in certain infant physical growth measurements than the mean values, because while the slopes between groups differ, the only between-group mean difference is in head circumference.

The variability most affected by the supplement is Rohrer’s Index, as this is the only measurement with highly significant inter-group differences in the slope of the sibling-sibling regression in both sexes. A high Rohrer’s Index could result from above average weight for average length, or from average weight for below average length. Likewise, a low Rohrer’s Index could result from above average length for

TABL

E II

I E

Reg

ress

ion

of b

irth

mea

sure

men

ts

of a

you

nger

sib

ling

(tr

eate

d vi

a m

ater

nal

supp

lem

enta

tion

pr

otei

n-ca

lori

e or

pla

cebo

sup

plem

ent)

on

thos

e of

an

olde

r si

blin

g (n

o tr

eatm

ent)

(a

= in

terc

ept,

b =

slop

e)

Mea

sure

men

t M

ales

Fe

mal

es

and

grou

p M

ean

(a)

f SE

. M

ean

(b)+

S.E.

M

ean

(a)+

S.E.

M

ean

(b)+

S.E.

R

esid

ual

r2

S.D

. R

esid

ual

SD.

- W

eigh

t (k

g)

A (

prot

ein-

calo

rie

supp

lem

ent “

) B

(pla

cebo

b,

Leng

th (

cm)

A

B Hea

d ci

rcum

fere

nce

(cm

) A

B R

ohre

r’s I

ndex

(w

t/13)

A

B Su

bsca

pula

r sk

info

ld (m

m)

A

B Che

st c

ircu

mfe

renc

e (c

m)

A

B

0.06

1.

09 +

0.3

7 0.

63 *

0.1

2 0.

31

0.37

2.

46 *

0.4

4 0.

25 +

0.1

4 0.

34

1.58

+0.

33

* 0.

52 k

0.1

1 *

0.31

0.

31

1.66

+ 0

.27

0.43

f 0

.09

0.27

0.

34

0.04

22

.92

+ 7.

07

0.52

kO.1

4 1.

81

0.22

0.

12

29.2

2 +

5.42

0.

40+0

.11

1.61

0.

21

37.7

6 f

7.69

32

.14

+ 6.

69

0.25

+ 0

.16

0.36

+ 0

.14

1.74

1.

73

27.7

1 f

3.35

0.

19*0

.10

19.3

3 +

3.44

*

0.43

+0.

10

* 0.

90

0.06

14

.50

+ 3.

60

0.55

kO.1

1 1.

06

0.36

1.

03

0.25

18

.39

f 2.

73

0.44

+ 0

.08

0.77

0.

38

2.66

+ 0

.41

- .0.

03 +

0.16

1.

48kO

.23

**

0.43

+ 0

.09

**

0.25

0.

00

2.67

+ 0

.35

- 0.

01 f

0.1

3 0.

20

0.00

0.

18

0.31

1.

5OkO

.28

**

0.42

+0.1

1 **

0.

20

0.24

3.24

f 0

.36

0.07

* 0

.09

0.70

0.

01

3.58

k 0

.38

- 0.

05 +

0.1

0 0.

78

0.00

2.

34 f

0.4

6 *

0.29

f 0

.12

* 0.

87

0.10

3.

70+0

&t

- 0.

05 +

0.1

1 0.

79

0.00

1.51

0.

05

18.5

5 +

4.05

0.

43 f

0.1

3 1.

48

0.20

1.

48

0.13

21

.22

+ 3.

30

0.34

+0.1

0 1.

28

0.19

26

.16k

4.10

0.

21*

0.13

21

.35

+ 4.

10

0.35

+0.

13

a n

= 55

(m

ales

), n

= 49

(fe

mal

es,

exce

pt f

or h

ead

circ

umfe

renc

e,

whe

re n

= 4

8).

b n

= 53

(m

ales

), n

= 50

(fe

mal

es,

exce

pt f

or c

hest

cir

cum

fere

nce,

w

here

n =

49)

. n

is t

he n

umbe

r of

pai

rs o

f si

blin

gs;

‘mal

e’

and

‘fem

ale’

ref

er t

o th

e se

x of

the

you

nger

mem

ber

of a

pai

r of

sib

lings

, th

e ol

der

mem

ber

bein

g of

eit

her

sex.

*

A-B

-gro

up

diff

eren

ce s

igni

fica

nt a

t P

-z 0

.11.

**

A

-B-g

roup

di

ffer

ence

sig

nifi

cant

at

P =

0.01

.

133

average weight, or from average length for below average weight. Conceivably, supplementation may have had a differential effect over these subgroups which may have remained undetected when the data for Rohrer’s Index were aggregated for all subjects together. The data suggested to us that Rohrer’s Index increased when the value of the first sibling was less than average. However, when the previous sibling was above average in Rohrer’s Index then there seemed to be a decrement in the index associated with the protein-calorie supplementation.

The birthweights and lengths of both siblings in the pairs of children that were deviating the most from the expected sibling-sibling regression in Rohrer’s Index were further investigated by defining two groups. In the first, the younger member of a sibling pair had a Rohrer’s Index whose residual was equal to or below the 15th percentile of that expected on the basis of the sibling-sibling regression in the placebo group. In the second, the younger sibling had a Rohrer’s Index whose residual was above or equal to the 85th percentile of that expected from the placebo sibling-sibling regression. The 15th and 85th percentiles of the residual distributions of placebo children were used to define the two groups, to exclude as much as possible any effects of random regression to the mean. Sexes were combined in this analysis because the sibling-sibling slope for Rohrer’s Index did not differ by sex within supplement groups (Table III). The first group comprised 23 children in A (supplement) and 16 children in B (placebo). The second group also comprised 23 children in A and 16 in B, although the samples were independent. Mean birth- weights and lengths are shown in Fig. 1 with their standard errors. The population means in Fig. 1 are the combined A and B group means presupplement. We are assuming that any of the means whose standard errors show no overlap are significantly different in the statistical sense (P < 0.05).

When the second sibling was below expectation in Rohrer’s Index (upper part, Fig. l), the first group A sibling (unsupplemented) had a mean birthlength signifi- cantly below that of the population mean, while the birthweight was close to that of the mean. Moreover, in this group the reason for a reduction in Rohrer’s Index after supplementation was a highly significant (P < 0.01) gain in birthlength amounting to an average 1.3 cm gain between 1st and 2nd siblings. There was a gain in length in the placebo group, but a less significant one as compared to group A (P < 0.06). Moreover, before supplementation, children in the A group had birthlengths signifi- cantly below those of their B group counterparts. In contrast, none of the anthropo- metric means differs significantly between A and B groups after supplementation, as the standard errors overlap. Birthweights declined from pre- to postsupplement; however, this was not significant and occurred both in the supplement and placebo groups.

When the treated sibling was above expectation in Rohrer’s Index (lower part, Fig. l), neither of the birth measurements of the first sibling differed significantly from the original population mean in either the supplement or placebo group. The reason for an increase in Rohrer’s Index after supplementation, was a highly significant increase (P < 0.01) in birthweight in group A (protein-calorie supple- ment). This increase amounted to an average gain of 238 g. This weight gain is not observed in group B (placebo). In both placebo and supplement groups, birthlength

134

decreases (P -C 0.05) pre- to postsupplementation. This decrease is of the same magnitude in the A and B groups, and is thus presumably not connected to supplementation.

A = Supplement B = Placebo

wt L wt L

(a) (cm) (Q) (cm)

g 200

100

Mean

-200 1

200 -

IOO-

- 100

Mean -

-1OO-

-2w-

1st Sibling High Rohrer Index

1st Sibling Low Rohrer Index

1st Sibling

??2nd Sibling

cm

1

+1

0

-1

1

+1

0

-1

Fig. 1. Mean birthweight and lengths ( f S.E.) of children whose Rohrer’s Index changed the most from first to second sibling. ‘1st sibling high Rohrer Index’ = 2nd child’s residual Rohrer’s Index below the 15th percentile of the distribution of residuals from group B sibling-sibling regression. ‘1st sibling low Rohrer Index’ = 2nd child’s residual Rohrer’s Index above the 85th percentile of the distribution of residuals from group B sibling-sibling regression.

135

Discussion

A major result of this study is that prenatal supplementation appears to have had a much more direct effect on the variance of birth measurements, particularly the covariance of siblings, than on the mean of these measurements, except possibly for head circumference. In Table III, when a regression slope was significantly different between groups, so was the intercept. This suggested that supplementation had equal but opposite effects on fetal growth at the extremes of the distribution of the birth measurements of a previous (unsupplemented) sibling. Whatever relationship existed between the birth measurements of siblings with adjacent birth orders, this has been disrupted by treating one sibling with increased calories and protein.

Usually a statistically significant increment in the mean of physical growth measurements in nutrition supplementation studies is taken as an indication of a treatment effect. Our data show that the infant index of body proportions (wt/13), and to a lesser extent birthweight, do increase when the birth size of a previous sibling is less than the average. However, our data also show that when the previous infant has an above average Rohrer’s Index, there is some decrease in the index and birthweight of a second infant associated with high calorie and protein supplementa- tion of the mother. At face value this latter result may seem undesirable from a nutritional viewpoint. However, as our analysis in Fig. 1 shows, the reason for a postsupplement decline in Rohrer’s Index in some families is an increase in birth- length relative to weight rather than in a decline in weight relative to length. Other studies have suggested using Rohrer’s Index as an index of nutritional status in utero and that ill health risk in infants is related to having a value which is too high or too low for the index [2,4,9,13].

The results suggest that children who are likely to respond are those with a prior sibling showing dysmorphic growth. This is particularly clear in the case of short length for weight (top, Fig. 1). It is probably also true of those with low weight for length. One is tempted to apply the well-known terms ‘stunted’ versus ‘wasted’ respectively, to the two classes above. The responders we have arbitrarily defined as those beyond the 15th and 85th percentiles of the distribution of Rohrer’s Index adjusted for the value of a previous sibling. These constitute about 46 infants (44%) of the protein-calorie supplement group. The mechanisms underlying these effects are not clear, except that their timing must be different. This can be inferred from the increased length of treated infants with an older sibling having a high Rohrer’s Index, and the increased weight in infants where the change in Rohrer’s Index from infant 1 to infant 2 is positive. Length and weight have their greatest growth velocity in the middle and at the end of pregnancy, respectively. Infants who are pro- portionately small (or large) in undernourished or well-nourished populations may represent adaptive responses to dietary conditions, which are under genetic or placental control. Dysmorphic children, on the other hand, may be indicative of failures in such adjustments. Environmental lability is thus more likely in these cases, and nutritional intervention in populations at risk may have some effective- ness. In any case, maternal nutritional supplementation does not affect all infants equally. Maternal physiology or varying conditions over pregnancy may limit the response to such interventions.

136

Acknowledgements

Dr. R. Quentin Blackwell was a principal co-investigator of this study during the period of data collection. At that time he was a staff member at the United States Medical Research Unit No. 2 in Taipei, Taiwan. His contribution in making this study possible is gratefully acknowledged. We thank the University of Texas School of Public Health for provision of computer time and Gay Robertson and the staff of the Word Processing Center for manuscript preparation. The Nestle Coordination Center for Nutrition, Washington, D.C., and The Ford Foundation supported in part the data analysis for this paper.

References

1 Blackwell, R.Q., Chow, B.F., Chinn, K.S.K., Blackwell, B.N: and Hsu, SC. (1973): Prospective maternal nutrition study in Taiwan: rationale, study design, feasibility and preliminary findings. Nutr. Rep. Int. 7, 517-532.

2 Brandt, I. (1979): Postnatal growth of preterm and full term infants. In: Human Physical Growth and Maturation, Methodologies and Factors, pp. 139-160. Editors: F.E. Johnston, A.F. Roche and C. Susanne. Plenum, New York.

3 Chow, B. (1973): Effect of maternal dietary protein on anthropometric and behavioral development of the offspring. In: Nutrition and Malnutrition: Identification and Management. Editors: A. Roche and F. Falkner. Plenum, New York.

4 Davies, D.P. and Beverley, D. (1979): Changes in body proportions over the first year of life: comparisons between light-for-dates and appropriate for dates term infants. Early Hum. Dev. 3, 263-265.

5 Herriot, R.M., Hsueh, A. and Aichison, R. (1978): Influence of maternal diet on offspring: Growth, behavior, feed efficiency and susceptibility. Final Report to A.I.D./C.S.D.-2944.

6 Lechtig, A., Habicht, J.P., Delgado, H., Klein, R.W., Yarbrough, C. and Martorell, R. (1975): Effect of food supplementation during pregnancy on birthweight. Pediatrics 56, 508-520.

7 Lee, J.A. (1980): Note on the interpretation of results of supplementation trials. Am. J. Clin. Nutr. 33, 333-337.

8 MacDonald, E.C., Pollitt, E., Mueller, W.H., Hsueh, A.M. and Sherwin, R. (1981): The Bacon Chow Study: Maternal nutritional supplementation and birthweight of offspring. Am. J. Clin. Nutr. 34, 2133-2144.

9 Miller, H.C. (1981): Intrauterine growth retardation, an unmet challenge. Am. J. Dis. Child. 135, 944-948.

10 Mora, J.A., deParedes, B. and Wanger, M. (1979): Nutrition supplementation and the outcome of pregnancy. I. Birthweight. Am. J. Clin. Nutr. 32, 455-462.

11 Pollitt, E. and Mueller, W.H. (1982): Maternal nutrition supplementation during pregnancy interferes with physical resemblance of siblings. Early Hum. Dev. 7, 251-256.

12 Rush, D., Stein, Z. and Susser, M. (1980): A randomized controlled trial of prenatal nutritional supplementation in New York City. Pediatrics 65, 683-697.

13 Woods, D.C., Malan, A.F. and Heese, H. de V. (1979): Patterns of retarded fetal growth. Early Hum. Dev. 3, 357-362.