During initial planning of the review, a protocol and conceptual framework were developed (APF) to itemize previously identified factors associated with adult obesity () and to guide the library search strategy and review process. These factors span three critical periods for children less than or equal to 5 years of age: prenatal (fetal development), infancy and early childhood. The review was not limited in scope to only biomarkers; and social determinants of health were also considered, including markers of socioeconomic status (SES) and food security. Gestational exposures such as smoking, fetal growth restriction, gestational diabetes and maternal weight gain for example were also considered. Furthermore, birth outcomes (birth weight, premature delivery), developmental characteristics (growth patterns, sleep patterns, IQ and cognitive development, and childhood obesity) and behaviours (breastfeeding, diet and television viewing) were all part of the framework.
Conceptual framework for the review.
Literature search strategy
A review protocol developed by one of the investigators (APF) served as a blueprint for the literature search strategy. In December 2009, a medical librarian (DS–see Acknowledgements) developed search strategies specific to the different databases, which included the following: MEDLINE (OvidSP, 1950 to November Week 3, plus in-process citations up to Dec 22, 2009), EMBASE (OvidSP, 1980 to 2009 Week 50), EBMR Reviews (OvidSP – includes Cochrane Database of Systematic Reviews, DARE, Central Register of Controlled Trials, Cochrane Methodology Register, Health Technology Assessment, NHS Economic Evaluation, and ACP Journal Club), and ISI Web of Knowledge (includes Web of Science [1899–present], Biosis Previews [1926–present], CAB Abstracts [1910–present], and Food Science and Technology Abstracts [1969–present]). The search was inclusive of all languages. In addition to the electronic searches, hand searches of key nutrition and epidemiology journals (e.g. American Journal of Clinical Nutrition, Obesity Reviews, Journal of the American Dietetic Association, American Journal of Public Health) were also completed. Citations were managed and organized with a citation manager RefWorks©.
The initial search strategy was conducted in MEDLINE, then adapted to fit the indexing and search interface functionality of the other databases. For specific search terms for MEDLINE as an example, see Appendix A. The search strategy included relevant subject headings and text words for obesity
combined with terms related to developmental exposures, including but not limited to breastfeeding, maternal smoking
, gestational diabetes
, feeding, growth
and socioeconomic status.
Broad terms related to early life and adulthood were also used in combination with obesity
text words to retrieve additional results and ensure that the search was comprehensive. The citations list was reviewed by four research assistants and three researchers. The PRISMA (formerly known as QUORUM) protocol was followed for accounting of the screening process (136)
Screening process and inclusion/exclusion criteria
The screening criteria were adapted based on the preliminary search results. At the outset of the review process, it was decided that only human research studies that included original research were to be included. In addition, only quantitative studies were included and qualitative studies (e.g. interviews, focus groups) and case reports were excluded. Genetic markers were not considered for this review; however, if other markers were also available in the same study, the citation was included. Markers of interest that were measured in children who were less than or equal to 5 years of age were included, and children and adolescents 6–17 years of age were excluded. For the adult measurements, only adults aged 18–50 years were included. A child and an adult measurement for the same individuals were required in all cases. Measurements of body mass index (BMI) or body composition (% body fat, waist circumference [WC] or waist-to-hip ratio [WHR]) were considered acceptable measures of adult body fat mass and obesity. Only healthy populations were included; studies were excluded if children had prediagnoses or existing health conditions, such as Down's syndrome, epilepsy, cancer or mental illness. The early marker data had to be measured in children who were 5 years of age or less. If, e.g. 5–9 year olds were investigated and followed into adulthood, this study would only be included if the data from the 5 year olds were presented separately. However, exceptions were made when the timing of parental data (e.g. education, employment and health status) were unspecified. These were assumed to be the same throughout the subject's childhood. All citations were reviewed independently by two reviewers. In cases where there was disagreement, the research team reviewed the studies and made decisions whether to include or exclude from the review.
All of the retained studies were abstracted independently by two research assistants. Discrepancies were resolved by a third research assistant or by team discussion. Detailed information related to the study characteristics was tabulated in an MSExcel (Microsoft Office, 2003) database. Specific study characteristics documented included citation information, document type, study type, definition of obesity, BMI data (e.i. measured or self-report), percentage body fat data (i.e. method of measurement, e.g. anthropometry, dual energy X-ray absorptiometry [DXA]), childhood marker of interest, sampling method, sample size, childhood and adult ages at time of measurement of key variables and proportion of males/females. Other study details documented included subject ethnicity, exclusion/inclusion criteria, indicators of bias such as percentage of subjects who completed the study or who were excluded from the study and sources of funding.
Quality of the studies
Ranking the studies by specific quality assessments (e.g. criteria specific to study design and statistical analysis) was beyond the scope of this review; however, a number of factors were considered in evaluation of the results from each of the retained studies. Statistical rigor was carefully assessed, including the type of statistics completed and if there was adjustment for confounding variables. The type of study was also considered (prospective vs. retrospective) with the former being considered more rigorous. Similarly, measured variables vs. self-reported variables were considered more objective and reliable. It was preferred to have a measure of adult obesity (BMI > 30 or upper quintile of % body fat); however, some studies reported only overweight (BMI > 25) or trends in adult BMI/% fat mass.
Detailed data on the statistical analysis performed to describe the association between the early marker of obesity and adult obesity were tabulated. Given the diversity in the study designs and statistical analyses, a variety of statistics was reported. Information documented included risk assessment (odds ratio [OR], relative risk [RR] and hazard ratio [HR]), related means and frequencies (analysis of variance [anova], t-tests, chi-square), and associations (regressions, correlations). A P value of <0.05 was considered statistically significant. If trends were reported (often designated as P > 0.05 but P < 0.10), they were not abstracted. If multiple analyses were completed, the values adjusted for the most confounders were abstracted.
For the purposes of this paper, ‘possible’ and ‘probable’ early markers of adult obesity were identified. All identified markers were from studies that met all the predetermined criteria. A possible marker was defined as one whereby there were 6–10 studies that reported a positive relationship (in >80% of the studies) between the early marker and adult obesity in either female or male offspring. A probable marker was defined as one whereby there were at least 6–10 studies that reported a positive relationship (in 100% of the studies) between the early marker and adult obesity in either male or female offspring.