The 165 women who had babies in the optimum birthweight range, 3,500 to 4,500 g, are referred to as the reference women and their median diet, summarized in Table 1, as the reference diet in the present and previous papers. This diet may be a logical starting point in developing nutritional recommendations, but subject to many reservations.

TABLE 1

Median daily nutrient intakes, 165 mothers who had babies in the optimum birthweight range 3,500 to 4,500g.

	daily nutrient intakes
energy	kilocalories joules	1983 8297
protein	g	73.2
fat total	g	90.2
carbohydrate	g	230.4
fibre	g	18.5
starch	g	114.7
monounsaturated fats total	g	28.7
oleic acid	g	25.0
polyunsaturated fats total	g	12.8
w6 fatty acids
total w6 fats	g	11.3
linoleic acid	g	10.98
arachidonic acid	g	0.170
w3 fatty acids
total w3 fats	g	1.28
alpha-linolenic acid	g	0.930
eicosapentaenoic acid	g	0.070
docosahexaenoic acid	g	0.130
p:s ratio		0.359
n-6:n-3 ratio		7.794
minerals
calcium	mg	920
chloride	mg	4151
copper	mg	1.53
iron	mg	11.8
magnesium	mg	260
phosphorous	mg	1286
potassium	mg	2959
sodium	mg	2643
sulphur	mg	647
zinc	mg	9.93
vitamins
vitamin A (ret eq.)	mcg	1000
retinol	mcg	498
carotene	mcg	1720
thiamin	mg	1.20
riboflavin	mg	1.86
niacin	mg	15.2
niacin equivalents	mg	83.4
pyridoxine	mg	1.45
cyanocobalamin	mcg	4.61
pantothenic acid	mg	4.25
biotin	mcg	25.6
folic acid	mcg	183
vitamin C	mg	83.8
vitamin D	mcg	4.58
vitamin E	mg	6.87

The first reservation is that Table 1 only describes the diet of 165 nonrepresentative women nearly all living around a particular hospital in Hackney, in the east end of London. British recommendations which could be used to adyise young women would have to be based on the diets of a representative sample of healthy women in all regions.

THE DELAYED CONSEQUENCES OF DIET

The diets of 28 mothers of low birthweight babies in the Hackney study are summarized in Table 2. The median intakes of these 28 mothers were lower for all nutrients than the reference intakes in Table 1. Many people have thought that low birthweight must be the result of women not eating enough during pregnancy and have arranged trials of supplementation. There have been many such studies. A randomized controlled trial of protein supplementation had little effect on birth outcome (14). The effects on the newborn of supplementation during the last two trimesters of pregnancy have been very modest, except on the newborn of poorer populations known to be suffering from poor nutrition (15). There is a long history of pregnancy in times of food shortage and famine showing that perinatal mortality, congenital malformation and low birthweight are associated with conception during food shortage rather than food shortage during the second and third trimesters (16). Previous studies suggest indeed that the associations between maternal nutrition and birthweight must have their origin around the time of conception and in the early days thereafter. Animal experiments show that low birthweight in offspring can be produced by damage to the germ cells of the mother before mating by chemicals, radiation or nutrient deficits (17). Low birthweight is associated with slow-timing ovulatory maturation before mating and slow-timing embryonic growth after mating.

TABLE 2

Median daily nutrient intakes, 28 mothers who had low birthweight babies < = 2,500g.

	daily nutrient intakes		intakes as percentage reference women in Table 1
energy	kilocalories joules	1697 7100	85.5 85.6
protein	g	64.5	88.1
fat total	g	74.1	82.2
carbohydrate	g	199	86.5
fibre	g	12.3	66.5
starch	g	97.2	84.7
monounsaturated fats total	g	24.3	84.7
oleic acid	g	21.3	85.2
polyunsaturated fats total	g	9.57	74.8
w6 fatty acids
total w6 fats	g	8.36	74.0
linoleic acid	g	8.15	74.2
arachidonic acid	g	0.140	82.4
w3 fatty acids
total w3 fats	g	0."0	72.7
olenic acid	g	0.820	88.2
eicosapcntaenoic acid	g	0.050	71.4
docosahexaenoic acid	g	0.115	88.5
p:s ratio		7.231	92.8
n-6:n-3 ratio		7.231	92.8
minerals
calcium	mg	775	84.2
chloride	mg	3237	78.0
copper	mg	1.17	76.5
iron	mg	9.82	83.2
magnesium	mg	203	78.1
phosphorous	mg	1055	82.0
potassium	mg	2513	84.9
sodium	mg	2078	78.6
sulphur	mg	584	90.3
zinc	mg	8.45	85.1
vitamins
vitamin A (ret eq.)	mcg	597	59.7
retinol	mcg	323	64.9
carotene	mcg	1609	93.6
thiamin	mg	0.94	78.3
riboflavin	mg	1.41	75.8
niacin	mg	11.7	77.0
niacin equivalents	mg	26.1	83.4
pyridoxine	mg	1.19	82.1
cyanocobalamin	mcg	3.81	82.7
pantothenic acid	mg	3.64	8S.6
biotin	mcg	20.5	80.1
folic acid	mcg	167	91.3
vitamin C	mg	98.3	117.3
vitamin D	mcg	4.33	94.5
vitamin E	mg	5.39	78.5

In the Hackney study the 513 women had been pregnant for about 12 weeks when their diet was recorded. The effect of pregnancy on diet has been the subject of studies in many countries. Virginia Beal of the University of Colorado undertook a classic longitudinal study covering 20 years, and showed that women change their diets very little on becoming pregnant (18). The diets of 95 women recorded before conception and then month by month until the end of pregnancy had an increase of 1.3 per cent in mean energy intake at the end of the third month of pregnancy compared with a standard error of the mean of 2.8 per cent. The correlation coefficient between the preconceptional energy intake and the energy intake at the end of the first trimester for the 95 women was + 0.91. Changes in average diet during the first trimester of pregnancy were very small. Some women do increase food consumption in early pregnancy, and Beal found a just significant mean increase of 3 and 4 per cent in calories in the 4th and 5th months. Hyperemesis caused some women to eat less in early pregnancy.

TABLE 3

Maternal daily nutrient intakes most highly correlated with birthweight in the lower birthweight range; 255 mothers of babies below, and 258 mothers of babies above median birthweight of 3,270g.

	mothers with babies < 3270g n=255		mothers with babies => 3270g n=258
	r	p	r	p
protein	0.238	&le 0.001	0.024	ns
energy	0.225	< 0.001	0.015	ns
fibre	0.212	&le 0.001	-0.028	ns
minerals
magnesium	0.253	&le 0.001	- 0.057	ns
iron	0.247	&le 0.001	- 0.079	ns
phosphorus	0.243	&le 0.001	- 0.005	ns
zinc	0.238	&le 0.001	- 0.007	ns
sodium	0.237	&le 0.001	0.052	ns
potassium	0.208	&le 0.001	- 0.068	ns
calcium	0.184	0.002	0.024	ns
vitamins
thiamin	0.200	&le 0.001	0.025	ns
niacin	0.198	&le 0.001	- 0.011	ns
pantothenic acid	0.186	0.002	0.050	ns
riboflavin	0.183	0.002	- 0.036	ns
folic acid	0.173	0.003	- 0.081	ns
pyridoxine	0.168	0.004	- 0.048	ns

In the Hackney study by the end of the first trimester, mean and median Central nervous system malformations intakes may be assumed to have been close to those of the habitual diet. Any changes must have been small compared with the 20 to 30 per cent differences between the nutrient intakes of the reference mothers and the mothers of low birthweight babies seen in Table 2. It is therefore reasonable to suppose that the diets measured at 12 weeks gestation in the Hackney survey were close to the habitual diets of the 513 women There is a case, therefore, for advice on the habitual diet of young women to be close to and compatible with the reference diet in Table 1 until such time as extension of this study elsewhere suggests modification. Practical advice has to discuss particular foods, and the diets of the Hackney mothers may be allocated to food groups.

MOTHERS FOOD CONSUMPTION BY MAJOR FOOD GROUPS

Mother's diet by major food groups is summarised in Table 4 for the mothers of the low birthweight babies and the reference mothers.

TABLE 4

Diet by major food groups of mothers of low birthweight < = 2,500g and optimum 3,500 2,500g birthweight babies.

Inadequate total food intake is seen to be one risk factor for pregnancy outcome. For this particular London population the importance of particular food groups as subsidiary risk factors declines as the table descends. Cereals, including breakfast cereals, and eggs and egg dishes at the top of Table 4 suggest the importance of breakfast. The position of meat and meat products at the bottom of the table suggests that low meat consumption is not a risk factor for many women in this London population. Low intake of dairy produce explains 28 per cent of the difference in energy of the diets of the two groups of mothers. The foods of higher nutrient density are generally nearer the top of the table.

IS THERE REALLY A CONNECTION BETWEEN MATERNAL NUTRITION AND BIRTHWEIGHT?

Whether or not nutrition is a factor of consequence in the causation of low birthweight is still a matter of debate and controversy. Witnesses to the Social Services Committee of the House of Commons in 1-9 said that there was no connection between poor nutrition and low birthweight and the Committee complained in its report at the "considerable difference of opinion amongst our witnesses regarding the importance of diet in pregnancy" (19). When commenting on the statistically significant associations between maternal nutrient consumption and newborn dimensions in papers on the Hackney study now published, some correspondents have dismissed our correlations as "not proving cause and effect", although it was stated that no causality was claimed. One correspondent wrote: "Causality between relatively low maternal intakes of certain nutrients and relatively low birthweights of their infants cannot be adduced from the data presented". David Hume, a British philosopher of the 18th century, was among the first to publish the conclusion that cause and effect can never be assumed from such observations. Herman Bondi, a contemporary philosopher of science, has written (20): "We must never say that the theory has been proved. We can only say it has stood a test successfully. If, however, such a test goes against the theory, then the theory has been disproved and one has to start again with a new theory." Karl Popper has explained the purpose of scientific observations (21): "Repeated observations and experiments function in science as tests of our conjectures or hypotheses, ie. as attempted refutations".

The hypothesis that maternal nutrition has no connection with birthweight is very easily refuted or shown to be improbable. The improbability of no connection is measured by the significance of correlations between maternal nutrient intakes and birthweight and Table 3 shows for one population that these correlations are highly significant. Correlations never prove cause and effect but provide very powerful evidence against wrong hypotheses. The mothers of babies of low birthweight are shown in Table 3 not to be randomly distributed among mothers at all nutritional levels but to be concentrated among mothers of poorer nutritional status. The multiple correlations in Table 3 do, of course, leave unclear the detail of the actual connection between maternal nutrition and birthweight, but throw serious doubt on the hypothesis that there is no connection or, indeed, no very serious connection. The value of all nutritional recommendations depends on the assumption that there is a connection between nutrition and health.

Significant correlations were found for many nutrients between women92s nutrient intakes and birthweight for all 513 women. Such correlation statements are based on linear relationships, but did the size of the baby increase linearly with nutrient intakes of the mother, so that the more the mother ate on average the bigger the baby? Further analysis of the data showed that above the median birthweight of 3,270 g this was not so. The higher the nutrient intakes the higher the birthweight only applied to the half of the mothers who had babies of birthweights under 3,270 g. In Table 3 the pairs of observations have been divided into two subsets, Cumulative averages of maternal intakes of thiamin one subset of 255 mothers who had babies with birthweights < 3,270 g and the other subset of 258 mothers with babies at or greater than 3,270g. For the smaller babies birthweight is seen to be significantly correlated with the intake of many nutrients, but for the larger babies there are no significant correlations at all. The hypothesis tested is that the intakes of mothers of lower birthweight babies were lower than the intakes of mothers of higher birthweight babies, making a one-tailed test of significance appropriate.

These results have implications for any dietary recommendations as there was a plateau of apparent adequacy covering at least half the mothers for all nutrients. In the birthweight range over 3,270g there was no evidence that any mother might have done better, in the sense that her baby might have been bigger or have had a larger head circumference, if she had had more of any nutrient. This statement is, however, qualified below when discussing prepregnancy weight. It is unlikely that this plateau began precisely at the median birthweight and the data may be analysed in other ways to see if the starting point can be determined more precisely. Of all the vitamins thiamin can be seen in Table 3 to have had the highest significance and may be taken as an example. The cumulative average curves for the maternal thiamin intakes are shown in Figure 3 beginning at the highest and lowest birthweights. The statistical significance of the difference between the two cumulative averages is plotted in Figure 4 using t-test and it is seen that the difference reaches a significance level below the median birthweight but only shortly below. These curves may be repeated for other nutrients showing that the maternal intake of some nutrients are becoming gradually and progressively lower than the reference intakes as the birthweight falls below 3,270 g. The multiple correlations show that many nutrients were concerned.

WHEN IS IT DESIRABLE TO HAVE RECOMMENDATIONS FOR A PARTICULAR NUTRIENT?

Nutritional recommendations are hypotheses, and recommendations for particular nutrients are based on the assumption that intake may affect health in general. The assumption that if a nutrient occurs sufficiently widely in foods then everyone is likely to have a high enough intake so that a recommendation is not necessary is difficult to defend. For example, the United Kingdom report on recommended intakes of nutrients published in 1-9 said (22): "Magnesium occurs widely in foods, particularly those of plant origin, as it is a constituent of chlorophyll; a dietary deficiency of magnesium is unlikely to occur in health".

Table 3 shows that the correlation of maternal magnesium intake and birthweight was not only significant, but more significant than that of any other nutrient. This result alone throws doubt on the hypothesis: A dietary deficiency of magnesium in the United Kingdom is unlikely. One of the authors therefore undertook a review to see whether there were other reports of an association of low magnesium intake and low birthweight (23). Studies were found from a number of countries from 1966 reporting an association between preterm birth, miscarriage, low birthweight and pregnancy hypertension with low magnesium intake. Furthermore, trials of magnesium supplementation have claimed a reduction in the prevalence of low birthweight. These reports on the effects of magnesium deficits are supported by animal experiments showing a range of congenital malformations and sterility associated with low magnesium intake. During the same period associations were reported between heart disease, raised blood pressure and low magnesium intake in the non-pregnant population. Several books have been published on the physiological effects of low and high magnesium consumption (24 -26). There is a good case for dietary recommendations for maternal magnesium intake in the United Kingdom. The United States recommendation for women of all ages is 300 mg/day and 450 mg during pregnancy. The maternal intake of the reference mothers in Table 1 of 283 mg is close to the recommendation of 300 mg.

There is a case for recommended intakes in anticipation of pregnancy of all the nutrients listed in Table

1. There is no United Kingdom recommendation, for example, for pantothenic acid, apparently on the strength of the following belief without reference to any source (22): "Diets in the U.K. usually provide 10 to 20mg which is more than adequate." The median diet of our reference women in Table 1 shows a median intake of 4.4 mg/day. A study from the University of California reported an average intake during pregnancy of 17 teenage girls as 4.7 +1- 1.4mg/day with a range from 1.8 to 7.2 mg/day and 14 girls postpartum with an average intake of 4.1 +/ - 1.1 mg/day with a range from 1.9 to 6.6mg/day (27). The contradictions in these Figures suggest that the absence of any recommendation for pantothenic acid may merit re-examination and this view is reinforced by extensive animal experimentation. Lefebvres in 1954 reported to the French Academy of Sciences (28):

"We have shown that in the pregnant rat a deficiency of pantothenic acid without any external signs can provoke resorption of the embryo and congenital anomalies. Some are obvious malformations: exencephaly, anophthalmia. Furthermore subcutaneous oedema has been observed and haemorrhage of extremities followed by necrosis."

An important conclusion from the animal experiments is that pantothenic acid deficiency can cause infertility and pathological reproduction without any other symptoms at all in the adult, suggesting that tissue concentrations need to be higher for satisfactory reproduction than for normal activity. This has been shown to be so for a wide range of animals including fish and insects. Such animal experiments establish hypotheses for men and women.

THE VARIABILITY OF INDIVIDUAL REQUIREMENTS

The Figures in Table 1 relate to a group of 165 women and not to an individual. The introduction to the American RDAs says that having found a healthy segment of the population the ideal method would be, "then to assess statistically the variability among the individuals within the group, and finally to calculate the amount by which the average requirement must be increased to meet the needs of nearly all healthy individuals." These calculations can be done using the data from the Hackney study and suggest that an increase of about 50 per cent in all the Figures in Table 1 might be necessary to include about 95 per cent of the women in the population at large. Does this, however, produce an answer that 0is useful? Can there be any good reason for recommendations to be defined so that they meet the needs of nearly all healthy individuals and are 50 per cent higher than the median requirement? Is it expedient to recommend that everyone, or groups of individuals, should consume 50 per cent more than the healthy individual with a median intake? It is arguably better to make the median the standard or norm and then recognize that half the population may have higher requirements for individual reasons such as an active outdoor life. A standard defined by the median intake of healthy persons in a population segment is most useful for catering. The same standard defined by the median is a best starting point for the counselling of individuals, until special requirements have been assessed. Nutritional recommendations are hypotheses that can be tested. Such recommendations will not be suitable for everyone and if adopted will involve some disappointments. For example if it is assumed that the diet of the reference mothers described in Table 1 is the recommended or standard diet, then the hypothesis may be tested that: Women consuming the standard or reference diet do not have low birthweight babies. There were 28 low birthweight babies ( < 2500 g) in the Hackney study and 27 had mothers with diets inferior to the reference diet, the majority with lower intakes of every essential nutrient. The one "disappointment" was a mother of a low birthweight baby recorded as eating the reference diet, but she was under 5 ft in height and had a body mass of 21 kg/m².

BIRTHWEIGHT AND NUTRIENT DENSITY OF MATERNAL DIET>

The mothers of low birthweight babies had diets that were not only lower in protein and energy than the reference diets but were low in nutrient density or quality. Mothers of the low birthweight babies are seen in Figure 5 to have had a mean daily fibre intake almost 30 per cent lower than the reference women in the Hackney study. Fibre intake was, however, highly significantly correlated with the intakes of the following nutrients in order of statistical significance, r in all cases being greater than 0.5 (p < O.OOl): magnesium, folic acid, thiamin, zinc, phosphorus, pyridoxine, potassium, iron and niacin. Any effects of fibre were inseparable from those of these nutrients in this survey. Ten of these micronutrients, 5 vitamins and 5 minerals, are seen to be 20 per cent or more below those of the reference diet. Nutrient density is associated with birthweight, and also with head size and birth length (6). Figure 5 suggests indeed that one or more of the micronutrients may be more important than calories or protein.

The probability of birthweight being below 2,500g is illustrated in Figure 6 for different intakes of 4 B vitamins. In each Figure the 513 mothers were divided into 5 approximately equal groups. It can be seen that about 40 per cent of mothers had daily intakes in excess of the median intakes of the reference mothers of 1.20mg for thiamin, 1.86 mg for riboflavin, 15.2mg for niacin and 1.45 mg for pyridoxine. While these median intakes of B vitamins may had a mean daily fibre intake almost 30 per cent lower than the reference women in the Hackney study. Fibre intake was, however, highly significantly correlated with the intakes of the following nutrients in order of statistical significance, r in all cases being greater than 0.5 (p < 0.00l): magnesium, folic acid, thiamin, zinc, phosphorus, pyridoxine, potassium, iron and niacin. Any effects of fibre were inseparable from those of these nutrients in this survey. Ten of these micronutrients, 5 vitamins and 5 minerals, are seen to be 20 per cent or more below those of the reference diet. Nutrient density is associated with birthweight, and also with head size and birth length (6). Figure 5 suggests indeed that one or more of the micronutrients may be more important than calories or protein. The probability of birthweight being below 2,500g is illustrated in Figure 6 for different intakes of 4 B vitamins. In each Figure the 513 mothers were divided into 5 approximately equal groups. It can be seen that about 40 per cent of mothers had daily intakes in excess of the median intakes of the reference mothers of 1.20mg for thiamin, 1.86 mg for riboflavin, 15.2mg for niacin and 1.45 mg for pyridoxine. While these median intakes of B vitamins may higher intakes are not only harmless, but benefited some mothers and would have benefited other mothers with babies of birthweights below the median. Nutrient requirements are interdependent. In particular circumstances an increase in magnesium intake can increase the efficiency of thiamin utilisation (23, 29). An increase in thiamin can increase the efficiency of utilisation of a limited supply of calories (30). There can be nothing final about a diet based on the median diet of apparently healthy women and diets of even higher micronutrient density may be shown in due course to benefit minorities without harm to the majority.

Low pregnancy weight is a risk factor for low birthweight and infertility (6, 16). This has often been studied and Figure 7, for example, is taken from the U.S. Collaborative Perinatal Study and shows the steady decline in the percentage of low birthweight babies as prepregnancy weight increases (31). Figure 7 is based on 1,058 women all around average height of 63 to 65 inches (160 to 165 cm). The correlation of prepregnancy maternal weight and birthweight in the Hackney study was 0.224 only slightly lower than in this American study. Maternal prepregnancy weight, estimated at the time of the dietary survey, was correlated with birthweight, birth head circumference and birth length. The average prepregnancy weight of the reference women with babies in the range 3,500 to 4,500 g was 63 kg (9 stone 13 lb) and of the mothers of low birthweight babies was 57.4 kg (9 stone 0 lb) in the Hackney study. The average body mass index (weight in kg divided by height squared in m) of the reference women was 23.7 kg/m², which is close to the average of a sample of American women in reproductive age of 24.0 kg/m². It is arguable that 24.0 kg/m² is not only the average but the best body mass index before conception. The body mass index of the mothers of the 28 low birthweight babies was 23.0 kg/rn2.

Prepregnancy weight and body mass index are associated with birthweight over the whole range of birthweights, unlike maternal nutrient intakes which were seen in Table 3 to have been significantly associated only over the range of birthweights below the median. It can be seen in Figure 8 that birthweight was significantly correlated with prepregnancy weight for the mothers and babies above the median weight, and for these mothers there was no significant correlation with calorie intake. For the mothers and smaller babies below the median birthweight the relative importance of prepregnancy weight and calorie intake was reversed; the association of birthweight and energy intake was significant, while prepregnancy weight was not significantly associated with either birthweight or energy intake. Kilocalories are the only nutritional variable shown in Figure 8, but if maternal intake of protein, essential minerals or B vitamins are substituted for kcalories, the importance of maternal nutrition in comparison with prepregnancy weight for mothers with babies below median weight is only emphasised. The mothers of the babies above median birthweight had a mean prepregnancy weight of 6l.3 kg and the mothers of the smaller babies below the median birthweight a mean prepregnancy weight of 57.7 kg. Prepregnancy weight is an important correlate of birthweight but, at least in the Hackney population of women, maternal nutrition was much more important than prepregnancy weight at the lower end of the birthweight range.

Prepregnancy weights reflect past diet. This is not necessarily the same as the diet consumed during the immediate prepregnancy period and early pregnancy. There was a negative correlation between prepregnancy weight and maternal intake of nearly all nutrients, indicating that thinner women were consuming more of all nutrients than heavier women. The thinner women had higher weight gains as would be expected from other studies. The heavier women had larger babies nevertheless. Prepregnancy weight and diet during pregnancy are not therefore independent variables. However for practical purposes there are two diets meriting consideration in preparation for pregnancy. There is the longer term diet necessary to achieve the best body mass index, and the desirable diet during the weeks and days before and around conception. Both of these diets are more important in determining birthweight than diet after diagnosis of pregnancy when it is too late to have a major effect on birthweight except in cases of extreme malnutrition (14-16).

BIRTH WEIGHT AND SMOKING

There were few significant differences between the nutrient intakes of smokers and non-smokers. Although the mean birthweight of the mothers that smoked was lower (3233 g vs 3302 g), the difference was not statistically significant in this study.

CONCLUSIONS: RDAs FOR WOMEN ANTICIPATING PREGNANCY

Recommended dietary allowances for women customarily include a category for pregnant women to take effect during the last two trimesters of pregnancy. A new category needs to be added for women in anticipation of pregnancy. The category might be called: "Women in anticipation of pregnancy or pregnant". The allowances for this category should be based on the median diets of women who produce healthy babies in the optimum birthweight range. Further studies should be pursued to estimate the habitual diet of a representative sample of women who have babies in the optimum birthweight range. It should be understood that if the median diets of healthy women are used for recommended dietary allowances, or standard diets, then adjustments will always be needed to meet the special requirements of individual women. Thus an increased allowance may be desirable for women with a body mass index below 24 kg/m² or with an energy expenditure above the median.

References

1. National Research Council. Recommended Dietary Allowances. (1980) United States, Washington DC, National Academy of Sciences.

2. Doyle, W., Crawford, M.A., Laurance, B.M. and Drury, p. (1982). Dietary survey during pregnancy in a low socio-economic group. Hum. Nutr. Appi. Nut,., 36A, 95-106.

3. Crawford, M.A., Doyle, W., Craft, IL. and Laurance, B.M. (1986). A comparison of food intake during pregnancy and birthweight in. high and low socio-economic groups. Pro gr. Lipid Res., 25, 149-54.

4. Doyle, W., Crawford, M.A., Wynn, A.H.A. and Wynn, SW. (1989) Maternal nutrient intake and birthweight. I. Hum. Nut,~. Diet, 2, 415---22.

5. Doyle, W., Crawford, M.A., Wynn, A.H.A. and Wynn, SW. (1989). Maternal magnesium intake and pregnancy outcome. Magnesium Res., 2, 205-10.

6. Doyle, W., Crawford, M.A., Wynn, A.H.A. and Wynn, S.W. (1990) The association between maternal diet and birth dimensions. .1. Nutr. Med., 1, 9-17.

7. Doyle, W., Crawford, M.A., Wynn, A.H.A. and Wynn, SW. The diets of 28 mothers of low birthweight babies. Midw.fe, Health Visitor and Community Nurse, forthcoming.

8. Paul, A.A. and Southgate, D.A.T. (1-8). McCance and Widdowson92s 91The Composition of Foods92, 4th edn. London, HMSO.

9. US National Center for Health Statistics. (1-2) A study of infant mortality from birth records. DHEW Publication No. (HMS) 72-1055.

10. Beck, L., Maier, F., Schmidt, E. and Rhode, J. (1-8) Mutter-und S~uglingssterblichkeit. Stuttgart, Kohlhammer.

11. Bakketeig, L.S. and Hoffman, H.J. (1983) The tendency to repeat gestational age and birthweight in successive births, related to perinatal survival. Acta. Ob.~tet. Gynecol. Scand., 62, 385-92.

12. Scottish Health Service (1-7). Scottish health statistics, 1-6, Edinburgh.

13. AlIman halso och sjukvard (1-6). Stockholm, Sveriges officiella statistik, 1-9.

14. Rush, D., Stein, ZA. and Susser, M. (1980). A randomized controlled trial of prenatal nutritional supplementation in New York City. Pediatrics, 65, 683-97.

15. Lechtig, A., Habicht, J.P., Delgade, H., Klein, R.E., Yarbrough, C. and Martorell, R. (1-5). Effect of food supplementation during pregnancy on birthweight. Pediatrics, 56, 508-20.

16. Wynn, M. and Wynn, A. (1981). The prevention of handicap of early pregnancy origin. London, Foundation for Education and Research in Childbearing.

17. Wynn, M. and Wynn, A. (1991) The case for preconception care of men and women. London, AB Academic.

18. Beal, V.A. (1-0). Nutritional studies during pregnancy. I Am. Diet. Assoc., 58, 312-20.

19. House of Commons. (1980). Second report from the Social Services Committee Session 1-9-1980. Perinatal and Neonatal Mortality. Evidence Vol.4, pp. 118-9 and Report Vol. 1, pp. 100-1. London, HMSO.

20. Bondi, H. (1-5). Herbert Spencer lecture 1-3. In Harre, R. ed. 91Problems of Scientific Revolution92, Oxford University Press, 1-5.

21. Popper, K. (1963). Conjectures and refutations. London, Routledge & Kegan Paul.

22. Department of Health and Social Security (1-9). Recommended daily amounts of food energy and nutrients for groups of people in the United Kingdom. London, HMSO.

23. Wynn, A. and Wynn, M. (1988). Magnesium and other nutrient deficiencies as possible causes of hypertension and low birthweight. Nuir. Health, 6, 69-88.

24. Seelig, MS. (1980). Magnesium deficiency in the pathogenesis of disease. New York, Plenum Books.

25. Durlach, J. (1988). Magnesium in clinical practice. London, John Libbey.

26. Wacker, W.E.C. (1980). Magnesium and Man. Cambridge, Mass., Harvard University Press.

27. Cohenour, S.H. and Calloway, D.H. (1-2). Blood, urine and dietary pantothenic acid levels of pregnant teenagers. Am. J. C/in. Nutr., 25, 5 12-7.

28. Lefebvres, J. (1954). Influence d92une d~ficience pantothenique leg~re sur les r~sultats de la gestation chez Ia ratte. CR hebdomadaires des seances de l92Academic des Sciences, 238ii, 2 123-5.

29. Itokawa, Y. and Kimura, M. (1982). Effect of magnesium deficiency on thiamin metabolism. Magnesium Bull., 3, 5-8.

30. Alvarez, OM. and Gilbreath, R.L. (1982). Thiamine influence on collagen during the granulation of skin wounds. J. Surg. Res., 32, 24-31.

31. Niswander, K.R. and Gordon, M. (1-2). The women and their pregnancies. DHEW Publication No. (NIH) 73-379. US Department of Health Education and Welfare.

32. Infant and perinatal mortality 1986 (1988). Birthweight, England & Wales. London, Office of Population Censuses and Surveys, Series DH3 88/1.

33. Congenital malformation statistics 198 1-85 (1988). England and Wales. London, Office of Population Censuses and Surveys. Series MB3 no. 2. (Received 12 July 1990; in final form 7 August 1990)

A.H.A. WYNN, M.A. CRAWFORD, WENDY DOYLE and S.W. WYNN