Article Text

Download PDFPDF

Preterm birth and subsequent insulin sensitivity: a systematic review
  1. Robert Tinnion1,
  2. Jenna Gillone1,
  3. Timothy Cheetham2,
  4. Nicholas Embleton1
  1. 1Department of Neonatal Medicine, Royal Victoria Infirmary, Newcastle upon Tyne, UK
  2. 2Department of Paediatric Endocrinology, Royal Victoria Infirmary, Newcastle upon Tyne, UK
  1. Correspondence to Dr Nicholas Embleton, Leazes Wing, Royal Victoria Infirmary, Queen Victoria Road, Newcastle upon Tyne NE1 4LP, UK; nicholas.embleton{at}ncl.ac.uk

Abstract

Objective The incidence of preterm birth is increasing worldwide. Evidence suggests that in later life these children are at increased risk of ‘metabolic syndrome’, which is itself associated with reduced insulin sensitivity (IS). We carried out a systematic review to examine whether preterm birth is associated with later changes in IS and whether a difference exists between those born small-for-gestational age (SGA) and appropriate-for-gestational age (AGA).

Methods We used the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidance to structure our review with a priori data extraction criteria to answer the questions posed and then carried out our literature search. Only papers which included preterm infants in their study population and specifically assessed IS were included. Findings are reported by age group to enable change over the life course to be examined, even though the studies were mostly cross-sectional, observation studies.

Results We identified and reviewed 26 suitable publications representing 20 separate cohorts, of which 16 had a term control group. The heterogeneity of the methods used to measure IS precluded meta-analysis. In infancy and early childhood there is a measurable association between IS and preterm birth. In later childhood and adulthood the strength of this association reduces, and current body composition becomes the variable most strongly associated with IS.

Conclusions There is an association between preterm birth and IS throughout the life course, but the data are conflicting and associations are likely to be affected by the heterogeneity of each study population and multiple confounding factors that may change over time. While the optimal nutritional strategy for preterm infants remains to be determined, standard public health guidance to avoid obesogenic lifestyle factors remains equally important to individuals born preterm.

  • Endocrinology
  • Metabolic
  • Neonatology
  • Nutrition
  • Infant Feeding
View Full Text

Statistics from Altmetric.com

What is already known

  • Being born small-for-gestational age increases the risk of later metabolic disease.

  • Preterm birth may disrupt nutritional programming ‘windows’.

  • Studies have arrived at different conclusions regarding the effect of preterm birth on subsequent sensitivity to insulin in adulthood.

What this study adds

  • Preterm birth is associated with measurably reduced insulin sensitivity at various stages of later life.

  • With increased maturity the influence of environment and diet (manifest as body composition) becomes more powerful.

  • The time course of this relationship is unclear and determination of causality requires further research.

Introduction

Approximately 5% to 10% of all births are preterm (<37 weeks’ gestation) with long-term survival rates in developed countries now greater than 50% for infants born at 24 weeks’ gestation.1 Neurocognitive impairment is the most important adverse outcome of preterm birth, but increasing data show that features of the ‘metabolic syndrome’2 ,3 are also more prevalent (w1, w2). Reduced insulin sensitivity (IS) is a key component of the metabolic syndrome2 and is associated with reduced glucose tolerance,4 hypertension (w3), hyperlipidaemia and disordered postprandial physiology5. IS is associated with obesity although the direction of causality is unclear.6

Low birth weight (<2.5 kg, LBW) is associated with reduced IS4 (w4, w5) and there are also data to suggest that early-life growth during sensitive windows in the prenatal and postnatal period may programme later metabolic outcomes.7 ,8 Infants born preterm will generally be LBW. There are few data to distinguish between the effects of preterm birth, compared to LBW at term, on later metabolism, and preterm and term groups may have experienced in utero growth restriction (IUGR). The majority of published studies are epidemiological and focus on adults born small-for-gestational age (SGA, typically birth weight <10th centile) who were born at term and were also LBW. The objective of this study was to determine whether preterm birth was associated with IS in childhood or adulthood, and the effect of being born preterm SGA compared to appropriate-for-gestational age (AGA).

Methods

Search strategy

We followed the Centre for Reviews and Dissemination (University of York, 2009) guidance for performing systematic reviews and the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidance (2009).9 Eligible studies were identified by searching electronic bibliographic databases (OVID MEDLINE, EMBASE, CINAHL, SCOPUS and PUBMED Central) and the Cochrane Database of Systematic Reviews. Search terms used for medical subject headings and keywords were: ‘prematurity’ OR ‘preterm’ or ‘neonate’ OR ‘neonatal’ OR ‘SGA/small-for-gestational age’ AND ‘Insulin resistance’ OR ‘IS’ OR ‘metabolic syndrome’. Results were limited to articles in English and human studies. Hand searching of review article references and studies from established preterm cohorts was undertaken. No specified year limit was applied as a search term. The search was last executed in January 2013.

Article selection

The search yielded 2206 articles including 98 papers from hand searching (figure 1). Initial screening excluded 2138 articles based on title, abstract and removal of duplicates and we then reviewed 68 articles in full text. Only studies using a direct measure of IS and including preterm infants were included in the final review: therefore, a further 36 studies and 6 review papers were excluded to leave a final total of 26 studies. Studies reporting data from individual cohorts in different publications were included for completeness as they represented longitudinal follow-up or used different subjects at different times from an original cohort.

Figure 1

Overview of article selection process for the review.

Data extraction

Prior to article selection, we determined the key data items for review: (1) inclusion criteria (gestation or weight); (2) number of children/adults studied (‘n’); (3) age at measurement; (4) measurement technique; (5) study design and (6) principal study conclusions. Features identified as strengths or weaknesses were recorded (see online supplementary appendix 1). Gestation and weight at birth were important considerations for a contemporary neonatal intensive care unit (NICU) population. The participant numbers in the studies allowed us to consider the relative weight of study evidence and age at assessment enabled stratification of our results. Measurement technique was key to inclusion in the review and study design helped determine quality of the study.

Methodological quality was formally assessed using a modified Newcastle-Ottawa Scale (NOS; table 1). The NOS allows quantification of the quality of non-randomised and cohort studies within the fields of population selection (relevance to preterm infants) and comparability of cohorts studied and the quality of the methods used to determine the outcome of interest (direct assessment of IS). Unlike use of a single reported ‘grade’ of evidence the NOS gives an overall impression of quality (maximum quality score of 12 stars) as well as an easily accessible breakdown of strengths and weaknesses in the specific areas outlined.10 Our NOS weighted the criteria areas as two stars for optimal practice, with one star for acceptable practice. Two authors (RT and JG) independently extracted data and scored the papers for quality. Disagreement was resolved by discussion and review (NE).

Table 1

Modified Newcastle Ottawa Score

We considered whether meta-analysis would be appropriate for our data set. There was significant heterogeneity in methodology used to quantify and assess IS in the selected studies. Combining results would have been inappropriate as there is no clearly accepted way of standardising between methods and no universal expression of either absolute IS results or change in IS. Many of the studies relied on physiological markers to make estimations of change in IS (eg, split pro-insulin concentrations) or postmeasurement modelling (eg, homeostasis model assessment (of insulin sensitivity) (HOMA2)) which precludes direct comparison. In addition, some of the methods used are not well validated for use in children against ‘gold standards’ such as euglycaemic insulin clamp studies. If we had considered only those studies with the same methodology for meta-analysis, it would have limited the scope of the study and not reflected the measurement of IS over the life course.

Results

Description of studies included in the systematic review

The selected studies investigated cohorts at different ages, published between 2000 and 2012, with participants from early infancy through to adulthood, mainly in resource-rich settings. The studies encompassed 20 unique cohorts. The median NOS for the selected studies was 8 (range: 6 to 10) from a possible 12 (table 2). Sixteen studies had a term control group4 ,11–25 and 3 studied individuals originally recruited into interventional trials.13 ,16 ,26 One follow-up study was interventional (a weight-loss programme)20 and one compared cohorts before and after a change in feeding practice.27

Table 2

Newcastle Ottawa Scores for selected papers (by year then ranking)

Different methods were used to determine IS. Thirteen used a variation of a glucose tolerance test (GTT; intravenous short-sampled or frequently sampled GTT (FSGTT)11 ,14 ,15 ,21 ,25 ,28 ,29; oral glucose load4 ,16 ,23 ,24 ,26 or milk bolus30) combined with insulin sampling and modelling to give a measure of IS. Three5 ,31 ,32 used hyperinsulinaemic euglycaemic clamp techniques. Others used measures of glucose metabolism such as fasting insulin and glucose, or 32–33 split pro-insulin, combined with a modelled estimation of IS such as HOMA. The presence and significance of altered IS changes with age,33 so the results are presented by age group: infancy (<2 years), childhood (2–10 years), adolescence (10–18 years), early (>18 years) and later adulthood (>35 years). These results encouraged us to develop a logic model (figure 2) outlining the changes seen over the life course.

Figure 2

Logic model demonstrating the changes in insulin sensitivity in infants born preterm, from birth to childhood.

Infant studies

Three studies investigated early postnatal life.27 ,29 ,30 Pittaluga compared cohorts before and after a change in protocol for postdischarge feeding27 and showed that altered nutrient composition (principally extra docosahexaenoic acid (DHA) and protein) resulted in lower fasting insulin levels at 2 years. Leipälä and Gray investigated preterm SGA status and IS and are discussed below.

Childhood studies (2–10 years)

One study where just 3.5% of the children were preterm showed a linear association between birth weight and IS.18 Two studies using HOMA modelling in 2–9-year-old children born preterm showed reduced IS12 ,17 and that preterm SGA infants with appropriate catch-up growth had greater IS than term SGA controls.12 This study also showed that preterm and term AGA groups had greater IS than the term SGA infants. Gestation and birth weight were associated with IS even after correction for body mass index (BMI).

Two studies used frequently sampled intravenous glucose tolerance tests (FSIVGTTs) in children 4–10 years old.11 ,15 Preterm children showed reduced IS that correlated with weight SDS at the time of testing, change in weight from term to 1 year and change in weight from term to current.11 The other study showed reduced IS in preterm compared to term controls and that those born preterm had similar IS to children born SGA at term.15 Increase in weight and height SDS and a higher weight-to-height ratio was associated with reduced IS.

Adolescent studies (10–18 years)

Fewtrell and Singhal13 ,26 followed up a cohort originally enrolled into nutritional trials in infancy.34 At 9–12 years of age,26 fasting insulin, glucose and pro-insulin were strongly related to current weight SDS. Regression modelling showed that change in weight SDS from 18 months to current age was positively related to insulin, pro-insulin and split pro-insulin. Birth weight had a strong negative correlation with 30-min glucose levels. By 13–16 years of age,13 a decrease of 13.4% IS per unit increase in weight SDS was seen. Adolescents randomised to higher nutrient intakes as neonates had greater split pro-insulin levels even after adjustment for potential confounders. A stepwise increase in adjusted 32–33 split pro-insulin was found to mirror quartiles of weight gain in the first two postnatal weeks.

Chan24 investigated 11–15-year-old children born term and preterm and conducted a standard oral glucose tolerance test (OGTT). They showed lower glucose levels at 2 hours postload in those born preterm. Reinehr20 studied SGA children 5–13 years of age (term and preterm) who were obese and assessed the effects of a weight loss intervention on IS. Birth weight had a small association with IS. Reduction in obesity was associated with improved IS.

Early adulthood

Three studies indicated that current body composition (especially high truncal fat) was the strongest determinant of reduced IS in adults born preterm.14 ,32 ,35 Rotteveel32 demonstrated that IS was not associated with perinatal factors after correction for current body composition. In a separate study on the same cohort they showed that IS was independently associated with height gain (from 1 to 5 years old) and weight gain (2 to 21 years old).31 Willemsen used FSIVGTT to compare IS and body composition measured by dual energy X-ray absorptiometry (DEXA).14 Adjusted IS was significantly influenced by height and weight SDS. Total fat mass and truncal fat mass were the most significant variables influencing IS. Birthweight SDS and gestation were not significant determinants after correction for fat mass. Later analysis of the same cohort suggested that growth patterns were important determinants of adult outcomes.25 Those with rapid infant catch-up growth had higher adult body fat percentage and waist circumference, but there were no significant associations with IS. When analysed by quartiles of weight SDS gain from birth to term, the highest quartile weight gain group had higher body fat, waist circumference, insulin response and disposition index although there were no associations with IS.

Finken35 showed that rapid weight gain to 3 months was associated with higher fasting insulin and lower IS in preterm born adults at age 19 years. However, adjustment for current body composition and other confounders resulted in loss of significance. There were also strong interactions between birthweight SDS, current fat mass or body fat percentage and HOMA index: having a higher fat mass after lower birth weight was associated with reduction in IS.

By contrast, Hovi4 reported a preterm cohort from Helsinki studied at 18–27 years of age and demonstrated reduced IS irrespective of BMI when compared to term controls. While the effect of prematurity on IS was not affected by adjustment for BMI, neither fat mass nor fat mass index was included in the modelling.

Pandolfi23 studied adults born LBW and showed preterm birth but not adult BMI to be a determinant of IS: they found reduced IS in LBW and preterm AGA even when corrected for current BMI, using BMI as a dichotomised variable in their modelling. Dalziel16 studied adults born preterm at age 30 years and found that 2-h IS after OGTT was reduced compared to term born controls. Birth weight adjusted for GA was not associated with reduced IS.

Insulin sensitivity in later adulthood (>35 years)

We did not identify any studies directly measuring IS after early adulthood. However, Kajante (w1) showed an association between the risk of developing type 2 diabetes and birth before 35 weeks’ gestation. A difference was also found between those born at term who developed type 2 diabetes and those who had reduced IS alone as measured by OGTT: while both were associated with markers of restricted fetal growth (LBW) and accelerated childhood growth (height), those with reduced IS were thin during childhood. Those who developed type 2 diabetes had a high childhood BMI.36

SGA status in infancy and childhood

Leipälä29 used FSIVGTT in preterm infants and found no independent effect of SGA status, although postnatal steroid administration was associated with reduced IS in the SGA infants only. Gray30 used a milk tolerance test in the first 2 months, showed that IS was related to weight at the time of test but not to gestation and that SGA preterm infants had higher insulin levels.

Five studies in early childhood showed no clear effect of SGA status on IS.11 ,12 ,15 ,17 ,37 Bazaes28 used HOMA modelling with IVGTT in 5–7-year-old children and found those born SGA had reduced IS. While Reinehr20 in a weight loss study in children 5–16-year-old demonstrated an effect of SGA status on IS, the study group included those born preterm and term. Chan24 found no difference between SGA/AGA groups in IS in early adolescence although those born preterm SGA had higher insulin levels 2 h after OGTT.

SGA status in adulthood

Rotteveel31 ,32 showed no effect of SGA status on IS after correction for adult body composition, although there was a difference in IS between SGA and AGA after adjustment for fat mass measured by bioimpedance. Hovi4 demonstrated that birth weight <10th centile in preterm infants, who then exhibited catch-up growth to term, had a 30% decrease in IS per weight SDS increase.

The figure shows the potential impact of factors over the life course. Note: arrow thickness varies based on the proportional influence of factors relevant to the evidence found in the studies included in this systematic review. The model should be read left-to-right. The term metabolic inflexibility is used to denote adverse changes in metabolism that may be developing as precursors to the metabolic syndrome.

Discussion

Our systematic review aimed to determine evidence outlining associations between preterm birth and later IS, and the effects of SGA status. Figure 2 demonstrates this using a logic model generated from our review. Logic models were originally proposed to examine complex systems where outcomes are not necessarily quantifiable using simple measures. Latterly they have been adapted as a tool for use in systematic reviews (w8) to allow either targeting of outcome measures or, as in our paper, a summary of the findings of the review in a flowchart. This allows the reader to follow changes through the system (in our case throughout life) from input to outcome. The logic model depicts the context we are interested in (ie, preterm birth), the care that is given (input) and the ways in which the effects of this input are measured (output). Importantly it also allows outcomes (ie, the summary evidence presented in the review) to be displayed at different stages with the magnitude of influence of factors associated with IS represented, using linear arrows, in the direction of influence.

In childhood, preterm birth is associated with altered IS.11 ,15 While weight catch up to that ‘expected’ based on parental size is associated with IS similar to those born at term in some studies,17 ,26 in others, increasing weight SDS was associated with decreased IS.13 ,18 ,26 By adulthood, the data are conflicting, with some studies demonstrating that fat mass is the major determinant of IS with no effect of gestation,14 ,32 ,35 while others identify a persisting effect of preterm birth.4 ,23 A recent meta-analysis did not conclude there were persisting effects of preterm birth on IS in adults alone.38

Our secondary outcome was to determine the effect of SGA status. Changes in the early postnatal period may simply reflect short-term homeostatic effects29 ,30 but the decreased IS observed in later childhood may be due to differences in early-life growth.4 ,28 ,31 ,32 However, other studies found little or no difference in IS related to SGA status.11 ,12 ,15 ,17 ,37 There are likely to be several reasons for the lack of agreement, including differences in methodology, population and definition of SGA: not all infants who are SGA will have experienced IUGR; similarly not all growth-restricted term infants will be SGA. Interestingly, glucose levels in AGA/SGA preterm groups were often similar.27 ,29 ,30

There are likely to be multiple mechanisms explaining associations between preterm birth and later IS. Preterm birth may be spontaneous or the result of a compromised pregnancy. Neonatal care in the first few weeks is complex, and recommended nutrient intakes are difficult to meet39 meaning many experience ex utero growth restriction, compounding pre-existing IUGR deficits. Early nutrition is primarily parenteral, using imperfect amino acid and lipid solutions and most receive a greater proportion of their calories from lipid, and lower intakes of protein, compared to the in utero fetus. Growth failure in early life may set the scene for later catch-up growth but a lack of controlled trials means that determining causality is difficult.

The physiological alterations determining the relationships between early-life events and subsequent IS are complex and may involve structural change within organ systems, alterations to endocrine feedback mechanisms (w2), cellular ageing and/or epigenetic mechanisms (w6, w7). Few of the studies reviewed adjusted for early-life factors such as nutrition and many did not adjust for childhood growth, obesity or lifestyle factors.

Only one controlled trial identified an association between more rapid weight gain in the first two postnatal weeks and decreased IS in adolescence. While current data suggest the possibility of associations with later epochs of growth, the data are conflicting and are open to bias, confounding and the possibility of reverse causation. Decreased IS may lead to increased obesity: equally, high body fat content may result in decreased IS. There is insufficient evidence to determine optimal nutritional regimens and whether these may differ for those born SGA.

Importantly, later lifestyle effects appear to be of greater significance than early-life exposures and continued efforts should focus on modifiable behaviours through childhood20 and into adulthood. However, lifestyle behaviours may be programmed by early-life events: preterm birth itself may alter later appetite or encourage more sedentary behaviour.40

This review is limited due to the heterogeneity in populations, early-life exposures, methodology of IS assessment, adjustment for confounders and the robustness with which current-life parameters have been assessed. Follow-up studies reporting outcomes in adults reflect neonatal care practices of 20–30 years ago, predating the widespread use of antenatal steroids, surfactant and parenteral nutrition, all of which have effects on survival and outcome. We endeavoured to provide a life course approach to IS, summarised in a simple model, while accepting that a review of cross-sectional studies at differing time points will not provide the same data as longitudinal studies, even after adjustment for any bias introduced by attritional losses over time.

Conclusions

There are associations between preterm birth and IS throughout the life course, but this is affected by multiple, confounding factors that change over time. Contemporary lifestyle factors confound this association and may be of greater magnitude. While the optimal nutritional strategy for preterm infants in early life remains to be determined, standard public-health guidance to avoid an obesogenic lifestyle is equally applicable to individuals born preterm. Future research must include prospective controlled trials with detailed measures of early exposures and longitudinal follow-up.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors All four authors are responsible for the reported research and have participated in the concept and design, analysis and interpretation of data, drafting or revising, and have approved this manuscript as submitted. RT conceptualised and designed the study, carried out the literature search, data extraction and quality scoring, drafted the first draft of the manuscript and edited the final manuscript for submission. JG carried out in parallel data extraction and quality scoring, and reviewed and revised the manuscript. TC reviewed and edited the manuscript, supervised the parallel data extraction process and contributed to the final submitted manuscript. NM had the original idea, provided ‘third-person’ arbitration during data extraction and quality scoring, and contributed to the final submitted manuscript. Dr Embleton is corresponding author. All authors have approved the submission of this version of the manuscript and takes full responsibility for it.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Atoms
    R Mark Beattie