Breastfeeding and feeding patterns
The breastfeeding rates reported in the present study were higher than among GDM patients in The United States  but were far from The Danish National Board of Health and WHO/UNICEF recommendations advocating full breastfeeding for the first six months of life [9, 10]. The observed rates were also considerably lower than normally reported in Denmark, the most quoted Danish survey  found less than 5 % of offspring not being breastfed at discharge, and the rate of fully breastfeeding at five weeks and at five months of age being 77 % and 32 %, respectively, corresponding to 8 %, 61 % and 18 % in our cohort. Recently, the median time of fully breastfeeding in Danish offspring was reported to be 122 days , while in our cohort the rate only had a comparable magnitude if fully and partly breastfeeding were considered together (Fig. 1). Complementary feeding was introduced early, 11 % even prior to the earliest official recommendation at the age of four months .
Previously, breastfeeding rates at hospital discharge have been shown to be lower after Caesarean than vaginal birth [23, 25], and recently it has been suggested that Caesarean birth prior to labour should be associated with the highest rates of breastfeeding difficulties . Present data supports this observation (Fig. 1). A labour-related endocrine process to initiation of milk production has been postulated as a biological plausible explanation for the breastfeeding difference , and counselling of mothers requesting Caesarean delivery without medical indication should include information on potential adverse consequences of Caesarean birth beyond the intrapartum period .
In general breastfeeding difficulties are well-known among women with GDM  and overweight mothers [9, 28], and nulliparous Danish women have been reported to have marginally lower breastfeeding rates than multiparous . Consequently, some degree of breastfeeding difficulty could be anticipated when the Danish surveys are compared and this is the reason for offering postpartum admission including extra breastfeeding counselling as mentioned in recommendations from The Danish National Board of Health . A specific procedure special to the children of the women with GDM is that they were all offered hypoallergic formula and a blood glucose test during the first hours following labour in order to avoid hypoglycemia, and Table 1 suggests an inappropriate effect from the early feeding regimen on breastfeeding initiation. Subsequently, the Danish national guideline on avoiding neonatal hypoglycemia has been fundamentally changed, but studies on the impact of the new guideline are still awaited.
The present group of women with GDM had received far more breastfeeding counselling than normal, and in spite of potential breastfeeding difficulties among this group of patients, we still consider lower breastfeeding rates compared with general Danish rates disappointing. This finding has not been reported previously.
As anticipated , assessment of anthropometric data was sensitive to choice of WHO or Danish reference standards, and it is debated which should be used in a Danish setting [9, 19, 20]. We analysed anthropometric data on the basis of both reference populations and mainly considered measurements clinically relevant if results were consistently significant. The most remarkable study finding was the significantly increased length and low BMI at the age of five weeks and at five months.
Larger studies have demonstrated prenatal exposure to hyperglycemia combined with maternal overweight conveying an increased risk of offspring overweight [13, 30], and we stratified offspring into groups exposed to maternal GDM and different degrees of overweight to investigate if this was the case in our smaller study. Differences between groups were insignificant but showed the same tendency as reported elsewhere.
Increased weight gain during the first four months of life has been reported as a strong, independent risk factor for childhood overweight in offspring from GDM patients, and preventing nutritionally-induced rapid early weight gain has been mentioned as a promising strategy to lower their long-term overweight risk . The mothers of the presented offspring cohort were in general well-regulated, and the low offspring BMI during the first five months of life is reassuring, but whether data should be considered the result of a successful intervention directed to the maternal life style during pregnancy, or whether it merely is affected by postnatal factors is unknown. The literature does not provide convincing evidence about a correlation between treatment of GDM and an improved offspring prognosis, but animal studies and human data strongly indicate that this could be the case . Further follow-up of the cohort during childhood will provide additional evidence concerning the obesity risk in the cohort. Although associations of maternal impaired glucose tolerance and increased offspring size at birth is well-known , we are not aware of other publications reporting isolated significantly increased length during the first months of living.
Anthropometric data and feeding patterns
There is an undisputed need to identify early determinants of obesity, and the association between breastfeeding and reduced risk of overweight and metabolic disturbances has been described as an outstanding example of the paradigm and concept of “perinatal programming” of health and disease . Offspring from GDM patients also seem to benefit from breastfeeding , and interestingly evidence is being gathered that breastfeeding may even have positive effects on maternal prognosis [33, 34], further emphasizing the importance of high breastfeeding rates among GDM patients. Low breastfeeding rates combined with early introduction of complementary feeding among the study cohort of offspring from women with GDM indicate an increased risk of later overweight , and the present study confirms an association of breastfeeding and low weight gain, but the overall study finding of low BMI at the age of five months, regardless of breastfeeding status, is surprising. However, the perspectives are complex and for instance the only large-scale randomized trial on the effect of breastfeeding has challenged the predominant conception of breastfeeding leading to a downward weight and length trajectory . Besides, a positive correlation between the neonatal intake of breast milk from diabetic mothers and later body weight has been reported , and increased concentrations of glucose and insulin, as well as a higher energy content of breast milk from women with diabetes, have been observed as compared with healthy mothers [1, 8, 36]. Potentially this might offer a dilemma to women with diabetes and their babies , and more research is needed to clarify whether breastfeeding might even have negative consequences with the risk of overweight and diabetogenic disturbances in the offspring of women suffering from glucose intolerance during pregnancy and lactation . However, in the present study, the women with GDM had been vigorously well-regulated approaching metabolic status of healthy pregnant women, and in clinical practice the alternative to breastfeeding is not banked human milk but formula. Considering the variety of advantages resulting from breastfeeding in general, breastfeeding should remain the preferred type of infant feeding in offspring from diabetic patients .
Data concerning the first five months’ of living from 99 % cohort of offspring from women with GDM is presented, and to our knowledge such a retention rate has not previously been reported in any other related study. However, there seems to be a dilemma between study particulars, cohort size and follow-up rate. Other well-conducted and larger follow-up studies on breastfeeding rates provided more detailed follow-up data but at the expense of the follow-up rate [1, 13, 25]. Studies reporting high retention rates faced selection bias including only the motivated families [37–39] at the study entry point. The impact of missing data on study conclusions is widely unknown, but it is reasonable to expect differences between families delivering follow-up data and non-responders. In the present study our priority has been to ensure data on the entire cohort, but a side effect was necessary restriction to available data on growth and feeding patterns from the examination by the general practitioner.
The mothers of the present cohort had GDM and were successfully managed on diet and other lifestyle interventions, making the intrauterine environment metabolically well-described. It is not known if results can be generalized to offspring exposed to more extreme intrauterine hyperglycemia or insulin treatment.
The reliability of measurements in the Danish primary health care system is difficult to assess. Correct measurements of head circumference and weight are simple, and regarding length, many general practitioners have infantometers or measuring boxes and it is mandatory that all length measurements should be done recumbent by a trained health care professional. There is no reason to suspect a systematic measurement error among the numerous general practitioners involved, but a tendency towards slightly overestimating length measurements cannot be ruled out. However, a systematic measurement difference between breastfed and formula fed offspring seems unlikely.
The high retention rate generated by the data from the general practitioners virtually excludes selection bias in the material; however selection bias affecting the comparison with the reference populations is still a possibility as the reference standard data have been derived differently.
While the association of breastfeeding and low BMI did have statistical significance, the question of causality cannot be assessed. The estimated associations may stem from reverse causation, i.e. that mothers adjust their feeding regimen according to observed growth. Besides, confounding is always a conspicuous risk in observational cohort studies. We were able to adjust for smoking and maternal pregestational BMI, but no data on socio-economic status were available. However, confounding from this source may primarily be mediated through smoking and obesity. Potential known and unknown confounders including paternal conditions and ethnic differences may interfere with results. To improve the clinical interpretability of the association analysis we chose to translate SD-scores to approximated kg/m2 (interpretation of Table 3, results section) even though this comes at a diminutive cost of strict mathematical precision.
The cohort consisting of 131 individuals is small when conducting analyses of associations or sub-groups. However, as a very rough rule of thumb, small studies imply more problems interpreting negative than positive results, and despite the study size, the analyses of association between breastfeeding and BMI reached statistical significance. Additionally, it is reassuring that previous reports of breastfeeding difficulties associated with maternal BMI and following prelabour Caesarean delivery seem to be confirmed. The study contributes to the existing literature but some subgroup analyses are too small for independent conclusions. Particularly, the partially breastfed groups may be quite heterogeneous, and the study has little power to assess optimality of mixtures of breastfeeding and other nutritional sources in the range between full breastfeeding and none.