Prenatal and Postnatal Milk Supplementation and Adult Insulin-like Growth Factor I: Long-term Follow-up of a Randomized Controlled Trial

The insulin-like growth factors (IGF-I and IGF-II) are peptides that have structural and functional homology with insulin. In addition to acute metabolic insulin-like actions, the IGFs mediate many of the effects of growth hormone on embryonic and postnatal childhood growth. The IGFs are present throughout the body, almost entirely bound to a series of six high affinity binding proteins (IGFBP-1 to -6; ref. 1). In the circulation, the majority of IGF-I is bound to IGFBP-3 and an acid labile subunit in a large ternary complex with restricted ability to cross the capillary endothelium. The ratio of IGF-I to IGFBP-3 is considered to reflect the bioavailability of IGF-I to target tissues (1). A number of recent prospective studies have shown that raised levels of IGF-I, low levels of IGFBP-3, and in particular a high molar ratio of IGF-I to IGFBP-3 are associated with increased risk from certain cancers (2, 3). A recent meta-analysis showed positive associations between raised IGF-I levels and premenopausal breast cancer, colorectal cancer, and prostate cancer (4). In contrast, low IGF levels may be associated with an increased risk of insulin resistance and type II diabetes (5), ischaemic heart disease (6), and cognitive decline (7).

Given the potential role of IGFs for a wide range of different age-related diseases, it is important to know whether environmental determinants may modify adult IGF-I levels. In particular, it has been suggested that adult levels may be influenced or “programmed” by intrauterine or postnatal growth (8). Observational studies that examine early life growth measures are confounded by childhood and adult social circumstances. This makes it hard to distinguish whether associations are related to a genuine critical or sensitive period effect (9). Long-term follow-up of randomized trials provides a unique opportunity to examine whether an early life intervention has a long-term influence, unconfounded by other factors. This approach has already been used to test the “fetal origins” hypothesis in relation to high blood pressure, insulin resistance, and markers of atherosclerosis (10-12). The aim of this study was to test whether milk supplementation given to mothers during pregnancy and to their offspring postnatally may have a long-term influence on adult IGF-I, IGFBP-3, and the molar ratio.

The Barry Caerphilly Growth study is the follow-up of an original randomized controlled trial involving pregnant mothers and their offspring, which continued until the age of 5 years. Pregnant mothers, between 1972 and 1974, were recruited through primary care from two towns in Barry and Caerphilly, in South Wales. They were randomized by an independent observer to either a “supplemented” or control group using random number tables. The supplemented group was provided with milk tokens throughout pregnancy and subsequently for their child (index case) until the age of 5 years, which entitled them to additional free milk delivered by their milkman. The women completed a questionnaire during pregnancy and the infants had regular anthropometric measures taken. Birth weight was obtained from hospital records and later weights and heights were measured by trained study nurses. Mothers were asked about dried milk consumption and this was quantified as ounces per day and then categorized into quartiles, as in a previous publication (13), so that exclusively breast-fed children would be in the bottom quartile. We examined feeding practice at 3 months of age. Between 1997 and 1999, we attempted to trace all subjects who had completed the original study (14). Subjects who agreed to take part in the follow-up completed a questionnaire and attended a morning clinic in a fasting status where they underwent standard anthropometric measures and blood sampling. Blood was spun at 3,500 rpm for 30 minutes and immediately stored at −20°C. Serum levels of IGF-I and IGFBP-3 were determined, blind to treatment status, by RIA, as previously reported (14). IGF-I, IGFBP-3, and ratio were transformed due to skewness (square root for IGF-I and ratio, and logarithmic for IGFBP-3) but back-transformed data are shown for ease of interpretation, having adjusted for other covariates. We examined baseline characteristics for all subjects that were initially randomized but for whom there was no IGF measure and for those with an available measure using logistic regression. To test for differential loss to follow-up, we examined for an interaction between a baseline variable and randomization status. We used multivariable linear regression models to test for differences between the intervention and control arm. We initially included control for age and sex. We then added any baseline variables that may have biased the results of which P value was <0.10. We finally added other potential adult confounding variables on the basis of prior knowledge from other studies rather than any ad hoc stepwise procedure.

The number of eligible subjects recruited and followed up at 5 and 25 years is shown in the flow chart (Fig. 1). From the original sample, 51% (633 of 1,251) of subjects provided a blood sample at 25 years which comprised 70% (633 of 951) of subjects who completed the trial at 5 years (for a more detailed description of the sample characteristics, see ref. 14). Certain variables, such as younger maternal age, unclassified social class at baseline, and lighter birth weight (see Table 1), predicted loss to follow-up. Five variables (maternal systolic blood pressure at first visit, maternal smoking at first visit, gestation, birth weight, and birth length) were identified that may have influenced differential follow-up. The crude mean values of IGF-I and its ratio were lower for subjects who were in the intervention arm (see Table 2). This difference was slightly enhanced after adjustment for age and sex (model 1). Stratifying the results by birth weight or mean parental height in tertiles made no difference to the results. We then added the covariates that may have influenced follow-up but they produced almost identical results (model 2). We finally added three adult covariates that we postulated a priori based on the current literature may have confounded adult IGF levels, but these, if anything, increased the difference across the intervention groups (model 3).

Our results show that individuals whose mothers were randomized during pregnancy to milk supplementation and who received this until the age of 5 years had lower levels of adult IGF-I and molar ratio in adulthood. These data are based on an intention to treat analysis from a randomized controlled trial. However, over the 25 years since recruitment, there has been reasonable loss to follow-up. Subjects who had measures of IGF differed systematically from the original sample although these differences were small in absolute terms and, other than birth weight, did not differ systematically across the two intervention arms. Furthermore, adjustment for baseline and adult covariates hardly altered the results, which suggests that our estimates are unlikely to be biased due to loss to follow-up.

There are several reasons why patterns of growth may be linked to IGF levels. Low birth weight infants have low levels of IGF-I (15). In children and young adults, however, levels are weakly inversely related to their birth weight, but differ in relation to patterns of postnatal growth (14, 16, 17). In childhood, low levels of IGF-I are also associated with short stature and low body mass index (18). Circulating IGF-I concentrations increase throughout childhood to mid-puberty and then slowly decline in adulthood (18). These patterns of association suggest that there may be critical or sensitive periods during the life course when diet and other growth-influencing exposures may have an important effect on short-term and possibly long-term levels of components of the IGF system.

Such observational associations could be explained by a simple genetic model whereby polymorphisms in one or more genes not only regulate IGF levels but also in turn determine the pattern of growth seen in individuals. Experimental evidence is therefore more powerful as, although not excluding genetic factors, this would show the influence of modifiable environmental factors.

We are only aware of one other study that provides experimental evidence in humans for a long-term effect of nutrition on the IGF axis. This comes from the Dutch famine, a natural experiment of extreme nutritional deprivation (19). In this small study of postmenopausal women, there was a linear trend so that increased exposure to the famine in childhood was associated with higher IGF-I and IGFBP-3. The authors postulated that extreme nutritional deprivation resulted in a resetting of homeostatic mechanisms so that there was a permanent overshoot after the famine. Our results support this hypothesis as supplementation, in our case, resulted in down-regulation of the IGF axis. This replication is important as it is possible that our findings may have occurred by chance despite conventional levels of statistical significance (20).

Paradoxically, two observational studies of Danish children found that increased protein and milk intake at 9 months (21) and 2.5 (22) years were associated with increased serum IGF-I, although no such cross-sectional association was seen at 10 years of age (21). A small randomized controlled trial of milk supplementation in 80 girls of age 12 years found a modest increase in IGF-I levels among girls who received the intervention after 12 months of follow-up (23). Unfortunately, we have no data on childhood diet other than randomization arm, so we cannot assess whether our differences in adult IGF-I levels are related to specific dietary components such as protein intake.

The differences we observe between the intervention and control arms is not large and one should be cautious in not overinterpreting our results. However, it is perhaps surprising that any effect has been shown with such a modest intervention. The IGF axis plays a critical role throughout childhood growth. Circulating IGF-I is primarily produced in the liver. Hepatic production of IGF-I is controlled not only by growth hormone from the pituitary (integrating central control) but also by insulin, and in addition many nutrients directly stimulate hepatic IGF-I production. There is then feedback control with IGF-I suppressing pituitary growth hormone output. An increase in nutritional input could therefore directly increase hepatic IGF-I, which then suppresses pituitary growth hormone. Our data and that from the Dutch famine study would suggest that there may be critical or sensitive periods in childhood when nutritional exposures could result in long-term resetting of the pituitary, programming adult IGF levels.

We cannot differentiate whether any effects of supplementation act in the prenatal or postnatal period or across both periods as milk was given both in pregnancy and postnatally until the age of 5 years. We favor the notion that any effects are postnatal in line with the “growth acceleration” hypothesis proposed by Singhal and Lucas (24). In this case, early accelerated growth would be associated with lower life course IGF levels and this may act as one of the biological pathways linking early growth with adult insulin resistance, hypertension, and possibly cardiovascular disease. Similarly, this pattern should be associated with decreased cancer risk and is consistent with the observational data on breast-feeding, which has been shown to be protective for IGF-related premenopausal breast cancers (4, 25). Understanding what modifiable effects influence life course IGF levels could have important consequences for future adult health.

We are very grateful to the subjects who participated in the original survey and who were willing to continue to be followed up in early adulthood. Bro Taf Health Authority helped with contacting the subjects. The original study was undertaken by the Medical Research Council Epidemiology Unit (Cardiff) funded by the Department of Health. The follow-up was funded by grants from the British Diabetic Association (1192) and British Heart Foundation (97020). The funders had no role in the design, collection, analysis, interpretation, writing up, or decision to submit this piece of work. Ethical approval for the study was given by the Bro Taf Local Research Ethics Committee.