Metformin in Gestational Diabetes: The Offspring Follow-Up (MiG TOFU): Body composition at 2 years of age

In the MiG trial, 751 women with GDM who required medication to control their hyperglycemia were randomized to either metformin or insulin treatment; their pregnancy outcomes have been reported (1). From two recruiting sites in Auckland, New Zealand, and one site in Adelaide, Australia, women who had consented to further follow-up were contacted by telephone at approximately the time of the child’s second birthday to explain the follow-up study and to confirm that they were still agreeable to participate. In Auckland, a home visit was arranged for the first part of the assessment during which maternal interviews and simple anthropometry measurements of the mother and child were made. A follow-up appointment was made within 1 to 2 weeks of the home visit to attend the Liggins Institute, University of Auckland, for the child to have a physical examination, a neurodevelopmental assessment, and a total body dual energy X-ray absorptiometry (DEXA) measurement. In Adelaide, women and their children were invited to the hospital and all the assessments were performed there. This follow-up study had ethical approval at each contributing site, and written informed consent was again obtained for each participant. The study was registered prior to its initiation under the Australian New Zealand Clinical Trials Registry (ACTRN12605000311651).

Questionnaires were completed by trained researchers, including assessment of the family’s socioeconomic conditions, home environment, any drug and alcohol intake, and health of the mother and child. Diet was assessed by 24-h recall and food frequency questionnaires. Usual activity of the children was assessed by a 24-h activity diary. The child underwent a general physical examination by a pediatrician and a neurodevelopmental assessment by a psychologist. The neurodevelopmental findings and detailed diet and activity assessments will be reported separately.

Anthropometry measurements of the mother and child included weight, height, leg length, head, chest, waist, hip and mid-upper arm circumferences, and biceps, triceps, and subscapular skinfolds. Skinfolds were performed with a Holtain skinfold caliper (Holtain Ltd., Crymych, U.K.). The method for each measurement was based on those used in a New Zealand Children’s Nutrition Survey (http://www.moh.govt.nz/moh.nsf/0/064234A7283A0478CC256DD60000AB4C/$File/nzfoodnzchildren.pdf) and detailed in the study manual. Training of study personnel was undertaken by a single person in Auckland to maintain consistency across sites. All measurements were repeated twice and the average calculated. A further measurement was made if the difference in measures was >0.5 cm (height and circumferences) or >0.5 mm (skinfolds), and the average of the two closest measures was calculated and used in the analysis.

Hand-to-foot single-frequency (50 kHz) bioimpedance analysis (BIA; BIM4, Impedimed, Queensland, Australia) of the child was performed with the child lying supine. Areas on the hand and foot where electrodes were to be placed were first cleaned with alcohol. The current electrodes were placed on the hand on the distal portion of the second metacarpal and on the foot over the distal portion of the second metatarsal. The sensing electrodes were placed at the anterior ankle between the tibial and fibular malleoli and at the posterior wrist between the styloid processes of the radius and ulna. The measurements were repeated up to three times until they were stable to within one ohm. The average resistance value was used in the prediction equation below. A BIA measurement was undertaken similarly in the mother, and the FFM was calculated as FFM = −9.53 + 0.69stature²/resistance + 0.17weight + 0.02resistance (9).

If consented to separately, a DEXA whole body scan of the child was performed on a Lunar Prodigy 2000 scanner (software version 4.80 × 6.50, General Electric, Madison, WI). Each scan was graded 1, 2, or 3 for quality by a single person in Auckland and a single person in Adelaide. Scans that were graded 3 (poor quality) were excluded from this analysis. As well as total fat, lean, and bone mineral content, an abdominal and thigh area for area fat content was calculated. Abdominal and thigh regions of interest were defined by the criteria of Ley et al. (10) The abdominal fat measure was obtained from analysis of a region positioned with the lower horizontal border on top of the iliac crest and the upper border approximately parallel with the junction of the T12 and L1 vertebrae. The sides of this region were adjusted to include the maximum amount of abdominal tissue. The thigh measure was obtained by analyzing an area of identical height placed over the thighs with the upper horizontal border positioned immediately below the ischial tuberosities. The lateral margins were adjusted to follow the shape of the thighs. The DEXA FFM was used as the criterion for the development of a prediction equation for bioimpedance FFM based on the following predictor values: weight, height²/resistance, sex (dummy coded with girls = 0 and boys = 1), and age. The equation developed was as follows:

(R² = 0.857, SEE [standard error of the estimate] = 0.559 kg), where H is height (cm), R is resistance (Ω), Wt is weight (kg), Sex (0 = girls, 1 = boys), and Age (months) (11).

The bioimpedance and DEXA measures of the child were performed in the morning before morning tea with the child wearing a T-shirt and dry disposable nappy.

A follow-up rate of 50% was anticipated, recognizing that it might be difficult, for various social reasons, to maintain contact with the MiG trial population, as others have described and highlighted by the initial 6–8 week postpartum follow-up, which was achieved in 75%.

A study of 240 children (120 in each arm) would allow detection of a 2% difference in body fat percent (based on an estimated body fat of 24 ± 4%) with 97% power and, thus, allowing a clinically meaningful analysis of the groups with respect to body composition. A study of 37 children in each treatment arm would have 80% power to detect a 2% difference in body fat percent.

For all data presented, the distributions were normal. Continuous variables are presented as mean ± SD. One-way ANOVA was used to test for differences in group means, and post hoc t tests were used to determine which groups were different. The significance level was set at 5%. ANCOVA was used to adjust for height, weight, and age when examining differences in fat mass and FFM among ethnic groups.

Of the women recruited into MiG at the two Auckland sites, 189 of 282 (67%) and 33 of 114 (28.9%) were seen for follow-up. In Adelaide, 101 of 181 (55.8%) were seen, giving a total of 323 women (Fig. 1). Body composition measurements were performed in 318 children, of whom 154 mothers and 164 mothers had been randomized to metformin and insulin treatment during pregnancy, respectively. A bioimpedance measurement was performed in 103 and 118 children in the metformin and insulin arms, respectively. DEXA measurements were performed in 140 children: 114 grade 1 and 2 scans were analyzed, 57 in each treatment arm.

The children seen at 2 years of age included a smaller proportion of those of Polynesian ethnicity compared with the total MiG population (14 vs. 20%, P = 0.02). Also, children seen for follow-up had had a shorter crown-rump length at birth (33.0 vs. 33.5 cm, P = 0.005) and smaller triceps skinfolds (4.80 vs. 5.15 mm, P = 0.0002) and subscapular skinfolds (4.95 vs. 5.20 mm, P = 0.07) at birth than the total group. All other baseline characteristics of the mothers and children and trial outcome measures were not different between the follow-up group and the total MiG population (data not shown).

In the children seen at 2 years of age, there were no differences between the groups in the baseline characteristics of the mother at randomization to treatment (Table 1). There were also no differences in pregnancy outcomes between the metformin and insulin follow-up groups, including the MiG trial primary outcome composite of neonatal complications (31.2 vs. 34.7%, P = 0.97), admission to the neonatal unit (17.5 vs. 18.3%, P = 0.97), and admission for >24 h (11.7 vs. 11.6%, P = 0.98). In addition, there were no differences between the groups in measurements at birth, maternal glucose control during pregnancy, and rates of breast feeding at 6–8 weeks postpartum (Table 2). Follow-up maternal anthropometry was not different between the two groups; maternal BIA showed higher FFM in the metformin group, but percentage body fat was not different (Table 3).

Body composition measurements at 2 years of age showed three significant differences (Table 3). The upper-arm circumference was larger in the metformin group (P = 0.002), and subscapular skinfolds and biceps skinfolds were bigger (P = 0.02 and P = 0.04, respectively). These results were explored further to confirm that the differences related to treatment. After adjusting for age, sex, ethnicity, and maternal glucose control during pregnancy, the P values were: upper-arm circumference, P = 0.005; subscapular skinfold, P = 0.01; and biceps skinfold, P = 0.02. There were no differences in DEXA measures between the two groups by unadjusted and adjusted analysis. This included total and regional fat measures. Bioimpedance measures also showed no difference between the metformin and insulin group in FFM or percentage fat.

This study describes the body composition in a unique population of 2-year-olds whose mothers had GDM and were randomized to treatment with metformin or insulin during pregnancy. The groups were matched for baseline maternal characteristics, maternal glycemia, and pregnancy outcomes.

Our initial hypothesis was that metformin exposure in utero would be associated with less central fat and, therefore, less insulin resistance in the offspring. However, we found no differences between groups in central fat measures, total fat mass, percentage body fat, or central-to-peripheral fat as measured by waist-to-hip ratio and DEXA-calculated abdominal-to-thigh fat ratios. Instead, we found that the children who were exposed to metformin in utero had larger upper-arm circumferences and bigger biceps and subscapular skinfolds. This suggests that exposure to metformin in utero has led to more fat being stored in subcutaneous sites, which may in turn mean there is less ectopic or visceral fat in these children.

These findings are important for two reasons: first, they suggest that maternal metformin treatment during pregnancy may lead to a more favorable pattern of fat distribution for exposed children; second, they suggest that simple measures of central fat may not be adequate for determining the potential effects of in utero exposure to metformin. The central fat measures used in this study provided a combined measure of subcutaneous and visceral fat, so further studies will be needed to confirm whether the children exposed to metformin have less visceral fat.

Size and location of fat cells are important predictors of insulin resistance and adverse metabolic consequences of obesity (4,8,12). Subcutaneous fat cells provide an important physiological store of extra nutrients. They have a limited capacity and are normally under homeostatic regulation, providing feedback about food intake and satiety. In situations of ongoing excessive nutrient intake, the adipocytes become large and dysfunctional and excess fat is deposited in visceral adipocyte depots, which readily release fatty acids and inflammatory adipocytokines (12). These changes are associated with insulin resistance, as opposed to insulin-sensitive obesity, which is associated with proportionally more healthy subcutaneous fat cells and less visceral fat (8,13,14). A more insulin-sensitive pattern of growth would be a plausible consequence of metformin exposure in utero, based on our understanding of metformin action (7). To examine this question further, ongoing follow-up will be important to determine whether differences persist and to measure visceral and subcutaneous fat and insulin sensitivity. Longitudinal follow-up is also important in that postnatal influences on growth may override any effect of metformin exposure during late pregnancy (15).

There are no other similar studies for comparison, so our data are novel. There are studies looking at subsequent growth of children whose mothers have had diabetes in pregnancy (16–20). Compared with children whose mothers did not have diabetes, they were more likely to be obese and have features of insulin resistance, which is felt to be the result of both genetic and intrauterine and postnatal environmental factors. It is possible that there are critical windows where intervention might improve these outcomes (21). There are two randomized trials showing that treatment of mild GDM (predominantly with diet) compared with standard pregnancy care was associated with improved pregnancy outcomes (22,23), but initial follow-up of children in one trial did not show a significant difference in BMI at 4–5 years of age (24). It is unclear whether the intervention in pregnancy was too late or inadequate or whether a difference will appear at subsequent follow-up. More detailed measures of visceral fat in those children would also be of interest. A further study has shown that treating women with GDM resulted in fewer overweight children at 5–7 years of age compared with children whose mothers had elevated glucose tolerance test results during pregnancy but did not reach the threshold for a diagnosis and treatment of GDM (25). These data also highlight the need for further studies looking at how different treatments for GDM influence long-term outcomes to better understand how to optimize the health of future generations.

The major strength of this follow-up study is that the offspring were well matched, enabling valid comparisons between treatment groups. Also, body composition was measured by several methods, and the differences found were consistent with a biologically plausible effect of metformin. A potential limitation is the low follow-up rate of the total MiG cohort. The follow-up group did have fewer Polynesian children, and as a group they had a shorter crown-rump length and smaller subscapular and triceps skinfolds at birth, compared with the total group. Otherwise, they were representative of the whole group and the study was adequately powered to explore differences in body composition. Also, additional analyses were performed to examine whether other potential confounders were contributing to the findings, but they confirmed that the differences related to treatment.

In conclusion, 2-year-old offspring of women with GDM, who were exposed to metformin in utero, had larger subscapular and biceps skinfolds but showed no difference in total or percentage body fat compared with children whose mothers were treated during pregnancy with insulin alone. Whether this will translate to a more insulin-sensitive pattern of growth requires further examination. The findings are reassuring for clinicians who are using metformin during pregnancy.

This study was supported by funding from the Health Research Council, New Zealand; the Auckland Medical Research Council; the Evelyn Bond Trust, Auckland; and the National Health and Medical Research Council, Australia.

W.M.H. was an invited speaker at the Merck European Association for the Study of Diabetes symposium on metformin in Stockholm, Sweden, September 2010. No other potential conflicts of interest relevant to this article were reported.

J.A.R. researched and interpreted data and wrote the manuscript. E.C.R. researched and analyzed data, contributed to the writing of the manuscript, and reviewed and edited the manuscript. V.O. analyzed data, contributed to the methods, and reviewed the manuscript. M.B., T.W., and W.M.H. researched data and reviewed and edited the manuscript.

Parts of this study were presented in abstract form at the International GDM Meeting, Pasadena, California, 9–11 April 2010; the International Society of Obstetric Medicine Meeting, Melbourne, Australia, 1–2 October 2010; and the Australasian Diabetes in Pregnancy Meeting, Sydney, Australia, 3–4 September 2010, as well as in talks without abstracts at the International Diabetes in Pregnancy Meeting, Salzburg, Austria, 23–26 March 2011; the Medical Complications of Pregnancy Meeting, London, England, 3–5 November 2010; the Lebanese Society of Endocrinology Meeting, Beirut, Lebanon, 13 November 2010; and the Reproductive Biology Meeting, Sydney, New South Wales, Australia, 29–31 August 2010.

The authors would like to acknowledge the additional people who performed clinical assessments and recorded data through the study: Aida Siegers, Jenny Rafferty, and Mariam Buksh from National Women’s Health, Auckland, New Zealand; Suzette Coat from the University of Adelaide, Australia; and Jewel Wen, Neil Snowling, Jennifer Crowley, and Sarah Bristow from the Auckland University of Technology, New Zealand.