Morrison 2008.

Methods	Study design: prospective cohort study. Analyses methods for cohort: regression model by stepwise selection from explanatory variables: age, BMI, IR and maturation stage at baseline; change in IR over 10 years' follow‐up; total calorie intake; percentage of calories from protein, fat and CHO (mean of interviews) during 10 years' follow‐up; and interaction terms (nutrients X baseline IR). How were missing data handled? NR. Number of study contacts: 10. Period of follow‐up (total period of observation): 10 years. Periods of recruitment: January 1987 to May 1988. Sample size justification adequately described? Reported for NGHS multicentre study. Primary consideration for sample size was adequate power for comparing change in subscapular skinfold between black and white girls. Sample size was increased to maintain sufficient power should only 65% of children be available for follow‐up measurements. Calculated target sample size was 1150 per group. Sampling method: convenient sampling by 3 clinical centres from public and parochial schools at Berkeley, Cincinnati and Westat (members of a medical program), USA. Study objective: to evaluate the role of preteen IR resistance and insulin in adolescent weight gain and the development of IFG and T2DM. Hypothesised that preteen IR, interacting with dietary factors such as total calories and fat calories, and 10‐year change in IR would positively predict 10‐year increases in BMI and the development of IFG and T2DM. Study population: white and black girls aged 9‐10 years living in Berkeley, Cincinnati and Westat, USA.
Participants	Baseline characteristics (reported as 1 overall group and 1 matched subsample) Overall Age (mean in years): overall (n = 639) 10.07 (SD 0.52); white (n = 280) 9.9 (SD 0.5); black (n = 359) 10.2 (SD 0.5); P < 0.001. Sex: 100% girls. Ethnicity: white, 43.8%; black, 56.2%. Parent education: high school: overall 22.5%; white 14%; black 29%; some college: overall 33.6%; white 18%; black 46%; college and beyond: overall 43.8%; white 68%; black 25%; P < 0.001. Parent income: household income: USD 10,000: overall 10.23%; white 2.6; black 16; USD 10,000‐USD 20,000: overall 10.07%; white 3.4; black 15; USD 20,000‐USD 40,000: overall 30.03%; white 26; black 33; > USD 40,000: overall 49.67%; white 68%; black 35%; P < 0.001. Pubertal stage: overall 58.5%; white 37.3%; black 75.1%; P < 0.001. Parental BMI: NR. Child total energy (kJ): overall (n = 521) 7517.85 (SD 1825.87); white (n = 241) 7439.15 (SD 1602.47); black (n = 280) 7585.59 (SD 1995.77); P = 0.68. Child total fat (%TE): Overall (n = 521) 35.07 (SD 5.1); white (n = 241) 34 (SD 5); black (n = 280) 36(SD 5); P < 0.001. Child total protein (%TE): Overall (n = 521) 14 (SD 3); white (n = 241) 14 (SD 3); black (n = 280) 14 (SD 3); P = 1.0. Child total CHO (%TE): Overall (n = 521) 51.93 (SD 7.07); white (n = 241) 53 (SD 7); black (n = 280) 51 (SD 7); P = 0.002. Child physical activity: NR. Child physical inactivity or screen time or both: NR. Child CVD risk (excluding fatness): HOMA‐IR: overall 3.12 (SD 2.74); white 2.45 (SD 2.29); black 4.64 (SD 2.94); P0.001; glucose (mmol/L): overall 5.2 (SD 0.41); white 5.17 (SD 0.44); black 5.22 (SD 0.39); P = 0.220; parents with T2DM: overall 9.1%; white 7%; black, 11%; P = 0.073. Child body fatness, BMI (kg/m2): overall 18.55 (SD 3.9); white 17.4 (SD 2.98); black 19.43 (SD 4.28); P < 0.001. Child body fatness, weight (kg): overall 37.91 (SD 10.54); white 34.2 (SD 7.7); black, 40.8 (SD 11.5); P < 0.001. Child body fatness, WC (cm): overall 64.78 (SD 9.04); white 62.2 (SD 7.5); black 66.8 (SD 9.6); P < 0.001. Subsample (paired matched at enrolment by pubertal stage, FM and insulin) Child body fatness, weight (kg), median: white (n = 172) 32.6; black (n = 172) 36.3; P0.001. Child body fatness, BMI (kg/m2), median: white (n = 172) 16.5; black (n = 171) 17.8; P0.001. Child body fatness, WC (cm), median: white (n = 167) 60.5; black (n = 214) 62.3; P0.001. Child body fatness, FM (kg), median: white (n = 172) 7.7; black (n = 172) 7.7. Child body fatness, % body fat, median: white (n = 172) 23.9; black (n = 172) 22.6; P0.01. Child CVD risk, median: HOMA‐IR: white (n = 143) 1.00; black (n = 168) 1.00; glucose (mmol/L): white (n = 143) 5.17; black (n = 168) 5.11. Included criteria: declared themselves as black or white; aged within 2 weeks of 9 or 10 years at time of 1st clinical visit; parents or guardians who identified themselves as same race as child; parents or guardians completed a household demographic information form and gave consent. Excluded criteria: other ethnic groups. Brief description of participants: 9‐ to 10‐year‐old black and white girls. Total number completed in cohort study: overall n = 639; white n = 280; black n = 359. Total number enrolled in cohort study: overall n = 2379; white n = 1166; black n = 1213.
Interventions	Description of exposure for cohort Time span: 10 years. Dietary assessment method: DR. Frequency of dietary assessments: single 3‐day DRs at baseline and during follow‐up (at 1, 2, 3, 4, 5, 7, 8 and 10 years). See Table 9; Table 10; Table 11; Table 12; Table 13; Table 14; Table 15; Table 16; Table 17; Table 18 for details of total fat intake exposure per outcome.
Outcomes	BMI BMI (kg/m2, 10 years' change). WC WC (cm, 10 years' change).
Identification	Sponsorship source: National Heart, Lung, and Blood Institute and the Lipoprotein Research Fund of the Jewish Hospital of Cincinnati. Country: USA. Setting: clinical centres (Berkeley, Cincinnati and Westat). Comments: NGHS. Author's name: John A Morrison. Institution: Division of Cardiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH (JAM); the Cholesterol Center, Jewish Hospital of Cincinnati, Cincinnati, OH (CJG and PW); the Department of Mathematics, University of Cincinnati, Cincinnati, OH (PSH). Email: john.morrison@cchmc.org; Glueckch@healthall.com. Declaration of interests: yes. "No conflicts of interest for any authors." No honorarium, grant, or other form of payment was given to anyone to produce the manuscript. "None of the authors had a personal or financial conflict of interest." Study ID: Morrison 2008. Type of record: journal article.
Notes	We contacted authors to request relevant regression data since they did not report the regression coefficients for total dietary fat intake alone as a predictor variable of body fatness in their regression models. We had not received a response by time of publication.
Risk of bias
Bias	Authors' judgement	Support for judgement
Were adequate outcome data for cohorts available? All outcomes	Low risk	Out of 639 girls with complete BMI outcome data, only 521 (81.5%) had dietary data. For 10‐year waist changes, 512 girls had complete data. No assessment comparing girls with dietary data compared to girls who did not.
Was there matching of less‐exposed and more‐exposed participants for prognostic factors associated with outcome or were relevant statistical adjustments done? All outcomes	High risk	Regression model (n = 521) performed by stepwise selection including age, BMI, IR and pubertal stage, 10‐year change in IR, total TE, percentage of calories from fat, protein, CHO during follow‐up period and interaction terms (nutrients × baseline IR). Physical activity/inactivity, parental BMI or SES not included in regression model. Secondary analyses (n = 172) with pair‐matched for race (black‐white); pubertal stage, BMI and insulin levels at 9‐10 years, adjusted for parental obesity level.
Did the exposures between groups differ in components other than only total fat? All outcomes	Low risk
Can we be confident in the assessment of outcomes? All outcomes	Low risk	Standard methods used for measurement of height, weight, skinfold and circumference measurements.
Can we be confident in the assessment of exposure? All outcomes	Low risk	Dietary intake assessed using repeated 3‐day DRs.
Can we be confident in the assessment of presence or absence of prognostic factors? All outcomes	Low risk	Data collection methods well described for most variables (e.g. pubertal staging, parental obesity).
Was selection of less‐exposed and more‐exposed groups from the same population? All outcomes	Low risk	All participants of the NHLBI growth and health study.