Abstract
Background:
Twin growth may be adaptively different from that of singletons to maximize survival in a shared intrauterine environment, translating to smaller average twin birthweight. We aimed to compare growth trajectories and overweight and obesity between twins and singletons to determine if twins ever catch up.
Methods:
We compared growth trajectories of twins (n=756) and singletons (n=30,264) from birth through 18 years in two national cohorts where children were measured every 2–3 years. We compared adjusted predicted means for each year of age using mixed models in each cohort and in the combined cohort. We stratified analyses by sex and evaluated overweight and obesity risk overall and at three stages based on average ages of puberty.
Findings:
Twins were significantly smaller than singletons for all measurements through 18 years, with differences in height ranging from 1.2 cm (95% CI 0.69, 1.8) shorter at age 3 and remaining 1.1 cm (0.10, 2.1) shorter at 18. Twins were 0.82 kg (0.49, 1.2) to 1.9 kg (0.46, 3.4) lighter between 3–18 years, with 0.10–0.76 kg/m2 lower BMI between 3–18 years. Findings remained when accounting for gestational age at delivery. Patterns were similar when stratified by sex, although girls may catch up in height by age 18. Twins had lower risk of overweight and obesity through early adolescence.
Implications:
Twins never caught up on height or weight but become more similar to singletons on BMI at 18. Lower risk of overweight or obesity in twins should be studied in the context of cardiometabolic health.
Funding:
This work received no specific funding awards but was supported by the intramural research program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
Introduction
Twin fetuses become smaller than singletons on average starting around 32 weeks’ gestation,1 with some evidence suggesting that intrauterine growth in uncomplicated twin pregnancies is adaptively different from that of singletons to maximize survival in an intrauterine setting of shared resources.2 Although twins are at increased risk for fetal growth restriction, it is unclear if growth restriction in twins confers the same long-term risk as growth restriction in singletons.2,3 Singletons with small-for-gestational age (SGA) birthweight, often considered a proxy for growth restriction, have higher risk of overweight and obesity in childhood and adulthood relative to appropriately-grown singletons.4,5 Considering twins’ differential fetal growth trajectory1 and smaller size at birth relative to singletons, there is a need to evaluate whether twin growth trajectories also differ in childhood. Determining whether twins catch-up in growth to singletons, or whether they are consistently smaller has important implications for cardiometabolic risk associated with SGA among singletons.6
Some longitudinal studies have found twins catch up at different ages up to age 6.7–10 Other cross-sectional studies have found smaller height, weight, and body mass index (BMI) in twins relative to singletons evaluated in adulthood,11–13 and adolescence,14 which may imply differences in growth. Previously,15 we found that twins remained smaller in early childhood (2–3 years) but may catch up in height to their singleton counterparts by 7–9 years. Other studies have more firmly concluded no difference in size between twins and singletons by age 416 or between the ages of 2–12.17 However, there are common limitations across all existing literature, including comparing twins to singleton references drawn from different populations to preclude direct comparison,7,8,16–20 lack of control for confounding factors,7,9,14,16–18,20 and cross-sectional analyses or limited age ranges for assessment of longitudinal growth or modeling accurate growth curves.7,11–18,20,21 Because patterns of growth may have different etiologies and implications for child health and development, longitudinal studies with repeated measures are needed. To our knowledge, no study has compared twin to singleton growth from birth to age 18.
In this study, we leveraged data from two large, nationally representative datasets that prospectively collected data from participants in the United Kingdom (UK) and the United States (US) from birth through adulthood. Our objective was to compare growth trajectories between twins and singletons to determine whether twins caught up on height, weight, or BMI. Additionally, we evaluated risk of overweight and obesity and whether risk was ever equal between twins and singletons.
Methods
Sample and anthropometric measurements
The Millennium Cohort Study (MCS), an ongoing longitudinal study, recruited participants born in the UK between September 1, 2000 and January 11, 2002, with data collection occurring at specified intervals.22 The first wave of data collection commenced when participants were age nine months, with follow-up interviews around age 3, 5, 7, 11, 14, and 17 years (y). Of the total cohort of 19,497 (excluding 11 sets of triplets),23 18,967 singletons and 502 twins (251 sets of twins) had at least one height or weight measurement.
Birthweight and the most recent child weight and date of measurement were reported at the first round of data collection, or “sweep” at 9 months. Information on birth length was not available. Beginning in the second sweep at 3y, child height and weight were measured using a standard scale and stadiometer by the interviewer. When the child refused or otherwise could not be measured, interviewers recorded the most recent height, weight, and date of measurement documented at a medical appointment (<1% of recorded measures, n=150/19,469). Height and weight data were cleaned by the MCS study team22 and cross-checked for this analysis against World Health Organization (WHO) and Centers for Disease Control and Prevention (CDC) sex-specific growth charts.24,25
The National Longitudinal Survey of Youth – 1979 (NLSY-79), is a nationally-representative and ongoing survey that began recruiting male and female participants in 1979 in the US.26 Beginning in 1986, NLSY Child and Young Adult (NLSY-CYA) was launched to follow the biological children of female participants (n=4,943) of NLSY-79, including a total of 11,545 child participants (excluding two sets of triplets). The total cohort with at least one height or weight measurement included 11,133 singletons and 246 twins (123 twin pairs). Participants in NLSY-CYA were directly interviewed every two years, beginning with the mother as the primary interviewee until the child was at least 14y.26 Data collection was scheduled at set 2-year intervals beginning in 1986 with the most recent data collection completed in 2020. The cohort was open to all children born to NLSY-79 female participants at any time, with the oldest child participant born in 1970 and the youngest born in 2008. This secondary analysis of public data was exempt from IRB review.
Birthweight was reported by the mother at the first interview that occurred after the child’s birth. Child height and weight were either directly measured by the interviewer or by the mother, under the supervision of the interviewer, using a standard scale and stadiometer. After 14y, participants directly reported height and weight to the interviewer. We checked height and weight data at each survey wave, with extreme values cross-checked against individual trajectories and WHO and CDC sex-specific growth charts.24,25
BMI for both cohorts was calculated as weight (kg)/height (m)2. We defined overweight and obesity using age (months) and sex-specific cut-points from the International Obesity Task Force.27 We used all BMI observations at any time between ages 3–18 and during three stages based on ages of typical pubertal onset:28 pre-puberty (girls=3–7y, boys=3–8y), early pubertal (girls=8–12y, boys=9–14y), late/post pubertal (girls=13–18y, boys=15–18y). Participants were categorized as overweight or obese if they had BMI meeting criteria for any measurement within the specified period of observation. The number of participants with height or weight information available varied by age for each cohort (Appendix p1), with good availability of measurements across ages (Appendix p2–4).
To evaluate the scale of differences over time, we additionally converted raw height, weight, and BMI to z-scores using WHO for ages 1–23 months and CDC for ages 2–18y, per CDC guidelines. Because twins are typically born at earlier gestational ages (GA) than singletons and WHO charts are based on term gestation for birth z-scores, we used the Fenton birthweight reference and corresponding LMS values to calculate all birthweight z-scores. We applied an age correction up to age 24 months,16 as recommended by WHO, to facilitate more accurate comparisons between children born preterm and term. Setting 40 weeks as term, we revised postnatal age for participants born prior to 40 weeks (e.g., postnatal age of 20 weeks but born at 36 weeks assigned corrected postnatal age of 16 weeks) before calculating z-scores.
Additional covariates were extracted from the earliest wave of survey data and supplemented by later waves, as needed. We included maternal age, which is associated with both twin status and fetal growth, as well as pre-pregnancy weight, height, BMI and child sex, which are strongly associated with child size. Child race/ethnicity was included to compare demographic characteristics only, as race/ethnicity classification was sufficiently different between cohorts to prevent data harmonization to adjust in growth trajectory models (Appendix p5). GA at delivery was reported by the mother, primarily, at the first interview after delivery. In MCS, GA was only reported in Sweep 1, 9 months after birth. For those who joined the cohort in Sweep 2, GA is missing (n=16 twins and 926 singletons). In NLSY, lag times were up to two years, depending on when the child was born relative to the interview, with some missing GA (n=42 twins and 1225 singletons).
Analysis
We compared maternal characteristics between twins and singletons for each cohort separately and in the full combined cohort, using ANOVA tests. We compared race/ethnicity between twins and singletons in each cohort separately using chi-squared tests. Further, we compared characteristics between the full sample and those with measurements at different observation times.
To compare twin to singleton growth trajectories, we fit linear mixed effects models with a cubic function for child age (months), random intercept and slope for age, and a nested intercept for family ID to account for twin clustering. We also included cubic splines at pre-specified knot points, based on natural inflection points from established growth charts and confirmed by loess plots of the data, where the shape of the height, weight, and BMI trajectory changed. Specifically, cubic splines were placed at 24 months for height, 24 and 84 months for weight, and 24 and 66 months for BMI.
We first fit models for twins and singletons separately for each cohort to ensure no difference in underlying growth patterns between children in the UK and US. In MCS we used all available data to fit weight models from birth through age 18. Height and BMI were modeled between 3–18y, since height collection began at 3y. In NLSY, we used all available data from birth to adulthood for height, weight, and BMI to model trajectories. Once we confirmed that trajectories were the same for each cohort, we combined the cohorts and repeated analyses, comparing marginal means for weight at 0, 6, 12, 18 months and every year between 2–18y and means for height and BMI at every year between 3–18y. All models were adjusted for maternal age, child sex, year of birth (categorized in five-year intervals), maternal pre-pregnancy weight, height, or BMI, depending on the outcome (e.g., adjusted for maternal height in child height models), and cohort to account for unmeasured differences between UK and US children. To better visualize changes in the magnitude of differences over time, we modeled height, weight, and BMI z-scores using similar linear mixed effects models. Given simpler trajectories of z-scores, these models differed in that they omitted cubic splines and substituted a quadratic function for the cubic function for age (months). Finally, we compared twin to singleton growth stratified by sex in the combined cohort.
We used Poisson regression models with a robust error variance to investigate risk of overweight and obesity between twins and singletons at specified time periods, using observed data. For Poisson models, we adjusted for the same variables as growth models, omitting child sex, as overweight and obesity cut-offs were sex specific. Statistical significance was determined using a two-tailed alpha of 0.05 for all comparisons.
We conducted a sensitivity analysis to account for potential long-term differences in growth trajectories that could be attributed to earlier GA at delivery for twins versus singletons, repeating all main growth analyses in the full cohort with an additional adjustment for GA. Observed differences in growth trajectories from this analysis can be considered independent of GA at delivery, with the limitation that all observations are dropped for participants with missing GA, reducing statistical power. To explore the robustness of our findings in relation to loss-to-follow up, we repeated main analyses among those who had at least one measurement in both: early childhood (<5y) and adolescence (>14y). All analyses were conducted in SAS v9.4 (SAS Institute, Inc., Cary, NC) or R Studio (R version 4.3.0, http://www.r-project.org).
Results
In both cohorts, maternal age was slightly older at the time of twin birth and GA at delivery was earlier for twins (36 vs. 39 weeks), but there were no appreciable differences in pre-pregnancy weight or height between twin and singleton mothers (Table 1). In comparing between cohorts, mothers in MCS tended to be older, with slightly higher pre-pregnancy weight and BMI than those in NLSY. MCS was also less racially/ethnically diverse. Compared to singletons, the prevalence of overweight or obesity was lower among twins at any time of assessment (Table 2). There were no differences in characteristics across twins or singletons between the full sample and those with measurements at later time points (Appendix p6).
Table 1.
Maternal and child demographic characteristics by twins and singletons in the UK Millenium Cohort Study (MCS) and the US National Longitudinal Survey of Youth (NLSY)
| Demographic | MCS | NLSY | ||||
|---|---|---|---|---|---|---|
| Overall (19,475) | Twins (502) | Singletons (18,973) | Overall (11,379) | Twins (246) | Singletons (11,133) | |
| Maternal age, years – mean (SD) | 28.3 (6.0) | 30.7 (5.6) | 28.2 (6.0) | 25.2 (5.9) | 28.2 (6.9) | 25.2 (5.9) |
| Maternal height, m | 1.64 (0.07) | 1.64 (0.08) | 1.64 (0.07) | 1.63 (0.07) | 1.63 (0.07) | 1.63 (0.07) |
| Pre-pregnancy weight, kg – mean (SD) | 63.7 (12.9) | 64.3 (12.8) | 63.2 (12.5) | 60.9 (13.1) | 61.3 (12.4) | 60.9 (13.1) |
| Pre-pregnancy BMI, kg/m2 – mean (SD) | 23.8 (4.6) | 24.3 (5.2) | 23.8 (4.6) | 23.0 (4.7) | 22.9 (4.2) | 23.0 (4.7) |
| Gestational age at delivery, weeks – mean (SD) | 39.3 (2.0) | 36.2 (2.8) | 39.4 (2.0) | 38.6 (2.3) | 35.9 (3.3) | 38.7 (2.2) |
| Child race/ethnicity, n (%) | ||||||
| Black | 742 (3.8) | 24 (4.8) | 718 (3.8) | 3065 (26.9) | 72 (28.9) | 2993 (26.9) |
| White | 15950 (81.9) | 436 (86.9) | 15,514 (81.9) | 3653 (32.1) | 86 (35.0) | 3567 (32.0) |
| Pakistani / Bangladeshi | 1355 (7.0) | 12 (2.4) | 1343 (7.1) | |||
| Hispanic | 2227 (19.6) | 36 (14.6) | 2191 (19.7) | |||
| Othera | 1410 (7.2) | 30 (6.0) | 1380 (7.3) | 143 (1.3) | 5 (2.0) | 138 (1.2) |
| Missing | 18 (0.09) | 0 | 18 (100) | 2291 (20.1) | 48 (19.5) | 2243 (20.2) |
BMI = body mass index, kg = kilograms, m = meters, SD = standard deviation
Other for MCS includes those identifying as Indian (502), Mixed (603), or Other (305); Other for NLSY includes those identifying as American Indian or Alaska Native (64), Asian or Pacific Islander (19), or Other (60)
Table 2.
Prevalence and adjusted risk ratio (aRR) of overweight and obesity between ages 3 to 18 years, UK Millennium Cohort Study and US National Longitudinal Survey of Youth combined cohort a
| Timing | Twin overweight, n (%) | Singleton overweight, n (%) | aRR (95% CI) of overweight |
|---|---|---|---|
| Any time | 255 (40.5) | 13,132 (49.3) | 0.78 (0.69, 0.88) |
| Pre-puberty b | 159 (25.7) | 8556 (33.7) | 0.67 (0.56, 0.80) |
| Early Pubertal c | 148 (29.6) | 7836 (36.1) | 0.80 (0.67, 0.94) |
| Late/Post pubertal d | 122 (31.4) | 6473 (35.2) | 0.92 (0.76, 1.11) |
| Twin obesity, n (%) | Singleton obesity, n (%) | aRR (95% CI) of obesity | |
| Any time | 105 (16.7) | 5406 (20.3) | 0.77 (0.61, 0.96) |
| Pre-puberty b | 67 (10.8) | 3147 (12.4) | 0.73 (0.54, 0.99) |
| Early Pubertal c | 46 (9.2) | 2669 (12.3) | 0.71 (0.51, 0.99) |
| Late/Post pubertal d | 45 (11.6) | 2476 (13.5) | 0.93 (0.67, 1.28) |
CI = confidence intervals
Overweight and obesity defined according International Obesity Task Force guidelines, available for 629 twins and 26,656 singletons with body mass index information
Pre-puberty state includes ages 3–7 years for girls and 3–8 years for boys
Early pubertal stage includes ages 8–12 years for girls and 9–14 years for boys
Late/Post pubertal stage includes ages 13–18 years for girls and 15–18 years for boys
Twins were consistently smaller than singletons at all points of observation in MCS for height, weight, and BMI (Appendix p7–8). Twins in NLSY were generally smaller than singletons but appeared to possibly catch up in their teens (Appendix p9). Differences in height were effectively zero by 13y, and although twins’ average estimates were lower through age 18, differences were no longer statistically significant starting at 14y for weight and 16y for BMI (Appendix p10).
In the full combined cohort of 748 twins and 30,106 singletons, twins were consistently smaller than singletons on all anthropometric measures between 3–18y (Appendix p11). Twins were consistently ~1 cm shorter than singletons between 3–18y (Table 3). Similarly, twins were one kilogram lighter at birth (1.01; CI 0.71, 1.31), becoming 2.02–244kg lighter between 14–18y. Twins also had about 0.50 kg/m2 lower BMI between 3–18y, although differences were not statistically significant at 18y. Patterns were similar for analyses using z-scores, with twins smaller than singletons in height, weight, and BMI (Figure 1), and there were marked differences in weight, narrowing across ages (Appendix p12), but remaining lighter at 18y. Differences in height z-scores also narrowed, with twins being shorter through 18y, although differences were not statistically significant at 17–18y, as in main analysis. In contrast to main analysis for BMI, twins remained significantly smaller in BMI z-scores through 18y.
Table 3.
Differences in mean height (cm), weight (kg), and BMI (kg/m2) between singletons and twins at each year of age through 18, UK Millennium Cohort Study and US National Longitudinal Survey of Youth combined cohort
| Age (years) | Height Difference (Singleton - Twin), cm | Weight Difference (Singleton - Twin), kg | BMI Difference (Singleton - Twin), kg/m2 |
|---|---|---|---|
| 0 | N/A | 1.00 (0.70, 1.31) | N/A |
| 0.5 | N/A | 0.79 (0.49, 1.08) | N/A |
| 1 | N/A | 0.70 (0.34, 1.06) | N/A |
| 1.5 | N/A | 0.69 (0.36, 1.01) | N/A |
| 2 | N/A | 0.72 (0.42, 1.02) | N/A |
| 3 | 1.22 (0.69, 1.75) | 0.82 (0.49, 1.16) | 0.63 (0.36, 0.89) |
| 4 | 1.24 (0.83, 1.65) | 0.96 (0.60, 1.32) | 0.45 (0.21, 0.69) |
| 5 | 1.25 (0.85, 1.64) | 1.12 (0.73, 1.51) | 0.49 (0.27, 0.70) |
| 6 | 1.24 (0.81, 1.66) | 1.30 (0.84, 1.75) | 0.58 (0.39, 0.78) |
| 7 | 1.21 (0.77, 1.66) | 1.48 (0.94, 2.02) | 0.66 (0.44, 0.88) |
| 8 | 1.18 (0.72, 1.64) | 1.67 (1.05, 2.29) | 0.71 (0.47, 0.96) |
| 9 | 1.15 (0.67, 1.62) | 1.85 (1.16, 2.53) | 0.75 (0.49, 1.01) |
| 10 | 1.11 (0.61, 1.60) | 2.02 (1.27, 2.77) | 0.76 (0.49, 1.04) |
| 11 | 1.07 (0.54, 1.60) | 2.16 (1.34, 2.98) | 0.76 (0.47, 1.06) |
| 12 | 1.03 (0.46, 1.60) | 2.28 (1.38, 3.18) | 0.75 (0.43, 1.07) |
| 13 | 1.01 (0.40, 1.62) | 2.36 (1.38, 3.34) | 0.72 (0.37, 1.07) |
| 14 | 0.99 (0.34, 1.63) | 2.39 (1.33, 3.46) | 0.68 (0.30, 1.06) |
| 15 | 0.98 (0.30, 1.66) | 2.38 (1.24, 3.51) | 0.63 (0.22, 1.03) |
| 16 | 0.99 (0.27, 1.71) | 2.30 (1.08, 3.51) | 0.57 (0.14, 1.00) |
| 17 | 1.02 (0.22, 1.83) | 2.15 (0.84, 3.47) | 0.51 (0.03, 0.98) |
| 18 | 1.08 (0.10, 2.05) | 1.93 (0.46, 3.40) | 0.44 (−0.12, 1.00) |
Figure 1.
Mean child z-score growth trajectories of singletons and twins, UK Millennium Cohort Study and US National Longitudinal Survey of Youth combined cohort
Trajectories of height, weight, and BMI z-scores estimated using linear mixed models with linear and quadratic terms for age (in months), and a random slope and intercept by age with a nested intercept for age by family ID to account for twin clustering. Models adjusted for maternal height (for height), maternal pre-pregnancy weight (for weight), and maternal pre-pregnancy BMI (for BMI), maternal age at delivery, and child sex
In analyses stratified by sex, results were similar to main analyses, with some sex differences in the last few years of observation (Table 4). Specifically, height differences between female twins and singletons began to diminish at 15y, becoming effectively zero at 18y, while a 1-cm difference in male height persisted through 18y. Weight differences of 0.82 kg for twin girls and 1.89 kg for twin boys remained at age 18y, though differences were no longer statistically significant between 16–18y in girls and at 18y for boys. A similar pattern was observed for BMI, where an approximate 0.50 kg/m2 difference remained for girls between 15–18y, although not statistically significant. For twin boys, BMI was 0.32–0.53 kg/m2 (−0.15, 1.22) lower at 17–18y relative to singletons.
Table 4.
Differences in mean height (cm), weight (kg), and BMI (kg/m2) between singletons and twins at each year of age, combined cohort, stratified by sex
| Age (yrs) | Height Difference (Singleton - Twin), Girls | Height Difference (Singleton - Twin), Boys | Weight Difference (Singleton - Twin), Girls | Weight Difference (Singleton - Twin), Boys | BMI Difference (Singleton - Twin), Girls | BMI Difference (Singleton - Twin), Boys |
|---|---|---|---|---|---|---|
| 0 | N/A | N/A | 0.97 (0.60, 1.35) | 0.99 (0.55, 1.43) | N/A | N/A |
| 0.5 | N/A | N/A | 0.72 (0.36, 1.08) | 0.73 (0.31, 1.15) | N/A | N/A |
| 1 | N/A | N/A | 0.61 (0.17, 1.05) | 0.62 (0.11, 1.13) | N/A | N/A |
| 1.5 | N/A | N/A | 0.60 (0.19, 1.01) | 0.61 (0.14, 1.07) | N/A | N/A |
| 2 | N/A | N/A | 0.64 (0.26, 1.02) | 0.65 (0.23, 1.07) | N/A | N/A |
| 3 | 1.15 (0.48, 1.82) | 1.16 (0.38, 1.94) | 0.75 (0.32, 1.18) | 0.76 (0.28, 1.23) | 0.79 (0.43, 1.15) | 0.36 (−0.04, 0.77) |
| 4 | 1.15 (0.60, 1.70) | 1.24 (0.63, 1.84) | 0.89 (0.42, 1.37) | 0.89 (0.37, 1.41) | 0.35 (0.02, 0.69) | 0.52 (0.17, 0.88) |
| 5 | 1.15 (0.61, 1.69) | 1.29 (0.70, 1.87) | 1.05 (0.53, 1.57) | 1.06 (0.49, 1.62) | 0.42 (0.12, 0.72) | 0.55 (0.23, 0.87) |
| 6 | 1.16 (0.59, 1.72) | 1.31 (0.68, 1.93) | 1.22 (0.61, 1.83) | 1.25 (0.58, 1.91) | 0.60 (0.34, 0.87) | 0.56 (0.28, 0.84) |
| 7 | 1.16 (0.57, 1.74) | 1.30 (0.65, 1.96) | 1.39 (0.67, 2.10) | 1.47 (0.67, 2.26) | 0.72 (0.42, 1.03) | 0.59 (0.28, 0.90) |
| 8 | 1.15 (0.56, 1.74) | 1.28 (0.61, 1.96) | 1.54 (0.73, 2.36) | 1.71 (0.80, 2.62) | 0.79 (0.45, 1.13) | 0.64 (0.29, 0.99) |
| 9 | 1.13 (0.53, 1.73) | 1.25 (0.57, 1.94) | 1.69 (0.78, 2.59) | 1.97 (0.96, 2.98) | 0.81 (0.45, 1.17) | 0.69 (0.32, 1.07) |
| 10 | 1.10 (0.49, 1.71) | 1.21 (0.50, 1.91) | 1.80 (0.81, 2.79) | 2.22 (1.10, 3.33) | 0.80 (0.42, 1.18) | 0.75 (0.36, 1.14) |
| 11 | 1.06 (0.42, 1.69) | 1.16 (0.42, 1.89) | 1.88 (0.80, 2.96) | 2.45 (1.22, 3.67) | 0.76 (0.35, 1.16) | 0.80 (0.38, 1.22) |
| 12 | 0.99 (0.32, 1.66) | 1.11 (0.33, 1.89) | 1.92 (0.74, 3.11) | 2.63 (1.28, 3.98) | 0.70 (0.25, 1.14) | 0.84 (0.38, 1.30) |
| 13 | 0.90 (0.20, 1.61) | 1.06 (0.23, 1.88) | 1.91 (0.62, 3.20) | 2.76 (1.29, 4.23) | 0.63 (0.14, 1.12) | 0.85 (0.36, 1.35) |
| 14 | 0.79 (0.06, 1.52) | 1.01 (0.15, 1.88) | 1.84 (0.45, 3.24) | 2.81 (1.22, 4.39) | 0.56 (0.04, 1.09) | 0.84 (0.30, 1.38) |
| 15 | 0.65 (−0.10, 1.40) | 0.98 (0.09, 1.88) | 1.71 (0.22, 3.20) | 2.76 (1.06, 4.47) | 0.51 (−0.05, 1.08) | 0.79 (0.21, 1.36) |
| 16 | 0.48 (−0.30, 1.26) | 0.97 (0.02, 1.91) | 1.50 (−0.09, 3.08) | 2.61 (0.78, 4.44) | 0.48 (−0.12, 1.08) | 0.69 (0.07, 1.30) |
| 17 | 0.27 (−0.59, 1.13) | 0.97 (−0.09, 2.02) | 1.21 (−0.50, 2.91) | 2.32 (0.32, 4.33) | 0.47 (−0.18, 1.13) | 0.53 (−0.15, 1.22) |
| 18 | 0.03 (−1.02, 1.08) | 0.99 (−0.32, 2.30) | 0.82 (−1.07, 2.72) | 1.89 (−0.40, 4.18) | 0.51 (−0.26, 1.27) | 0.32 (−0.50, 1.13) |
Risk of overweight for twins was 22% (aRR=0.78; 95% CI 0.69, 0.88) lower and risk of obesity was 23% (aRR=0.77; 0.61, 0.96) lower than singletons at any time between 3–18y. Results were similar for pre-pubertal and early pubertal stages (Table 2). There were generally no differences in risk between twins and singletons in the late/post pubertal stage (aRR=0.92; 0.76, 1.11 and aRR=0.93; 0.67, 1.28 for overweight and obesity, respectively).
Results of sensitivity analyses were similar to main analyses. After adjusting for GA at delivery, differences between twins and singletons were attenuated but statistically significant at the same ages (Appendix p13). Among participants with at least one measurement at both early childhood and adolescence, differences were similar in magnitude for height, with loss of statistical significance at 18y (Appendix p14). For weight and BMI, differences were slightly higher in magnitude than main analyses.
Discussion
In this large, prospective analysis combining two national datasets from the UK and US, we found that twins were consistently smaller than singletons in height, weight, and BMI from birth through late adolescence. These findings were robust in multiple sensitivity analyses accounting for differential gestational age at delivery between twins and singletons, as well as loss-to-follow-up. Twins also had lower risk of overweight and obesity than singletons at any time between 3–18y, and in pre-pubertal and early pubertal stages, prior to 13y for girls and 15y for boys. More similar risk of overweight and obesity between twins and singletons later in adolescence, as well as non-significant differences in BMI at 18y indicate that twins may eventually catch up to singletons on BMI. However, catch-up in BMI may simply reflect twins being proportionally smaller than singletons, given twins’ shorter stature and lighter weight through 18y. In sex-stratified analyses, girl twins appeared to catch up in height at 18y, while boy twins remained 1 cm shorter than singletons, which may reflect later puberty and a subsequently longer growth phase among boys. Similarly, twin-singleton differences in weight began to decrease for girls beginning at 16y, while boys remained 1.9 kg lighter at 18y. Girls and boys may catch up to singletons in BMI by 18y.
To our knowledge, no other study has evaluated twin growth longitudinally through age 18 using a singleton comparison group drawn from the same population as twins. However, our results are similar in magnitude to those of cross-sectional studies that directly compared twins to singletons from the same population between 16–25y. Two Norwegian studies found that male twins remained 0.6 cm shorter with 0.46 kg/m2 lower BMI, compared to singletons who were primarily 18–19y.11,12 When comparing twins to their singleton brothers, these studies found even larger differences of 0.9 cm shorter height and 0.58 kg/m2 lower BMI. Another Swedish study also found that male twins were up to 0.9 kg lighter, with small differences in height (0.4 cm) and BMI (0.2 kg/m2), unadjusted for potential confounders.13 Another longitudinal study comparing twins recruited from Japanese schools to an older singleton cohort selected from different schools, found that twins remained 2.6 cm shorter than singletons by 20y.9
Twin-singleton differences persisted in z-score analyses through age 17 for height, and through age 18 for weight and BMI, in contrast with some studies. These studies compared twins’ z-scores for height, weight, and BMI to reference populations of singletons,17 or country-based singleton child growth charts.7,8,16,18–20 In these studies, twin z-scores were compared to zero, representing the mean height, weight, or BMI for singletons. However, even country-based singleton references may not be directly comparable to twins.29 One longitudinal study found that twins from the Netherlands Twins Register (n=368) caught up to the Dutch singleton growth chart by 18y.19 However, in additional analyses comparing twins to their singleton siblings, the authors found that singleton siblings had higher BMI (z-score difference=0.31) relative to the Dutch singleton growth chart, and twins were shorter, lighter, and had lower BMI than their siblings by 18y. These inconsistent findings from the same study demonstrate the importance of comparing twins and singletons drawn from the same population to ensure comparability.
Despite limitations of existing literature, most studies have found that twins were smaller than singletons at various ages in childhood.7,8,14,15,18,19 We also observed differences by sex, which is consistent with others’ findings that girl twins catch up at ages when boys remained smaller.14,18,20 One cross-sectional study also found that female twins were lighter in weight but similar in height to singletons at age 45+.21 Regardless of whether twins catch up, their growth curves differ from those of singletons and our findings, combined with other studies, support the need for further evaluation of BMI cut-points to define overweight and obesity among twins. Twins were only slightly smaller than singletons in the current study, but height and weight define BMI and subsequent overweight and obesity. It is difficult to contextualize the meaning of finding lower risk of overweight and obesity in twins when BMI, on average, was lower in twins and overweight and obesity are based on singleton standards for BMI. Twin-specific pediatric growth charts may be warranted, considering the differences in twin and singleton curves and consequent implications for risk stratification related to overweight and obesity, as well as how growth percentiles are generally classified in twins.
There are multiple interpretations of our findings. Persistently smaller twin size through adolescence could be indicative of permanent in utero changes due to either pathological or physiological “growth restriction.”2 In the context of fetal growth literature and findings of smaller twin fat and lean thigh and abdominal growth that begins early in gestation,30 twins may be evolutionarily programmed for smaller size in preparation for a constrained intrauterine environment to maximize both fetal and maternal survival. Whatever the reason for persistently smaller twin size, future research should evaluate twin risk of adverse cardiometabolic outcomes in relation to height, weight, BMI, and related risk for overweight and obesity based on singleton standards.
A significant strength of our study is having longitudinal data from national cohorts that include both twins and singletons drawn from the same populations using the same selection criteria. This is an advantage over using data from twin registries, where participants volunteer to be involved in the registry and there are no singletons included for direct comparison. Furthermore, height and weight measurements were collected prospectively from birth or age 3 through 18y, which allowed for modeling of growth curves. Combining the cohorts also yielded a relatively large number of twins (~750), to provide adequate statistical power to compare twins to singletons. By adjusting for gestational age and applying age correction to z-scores, we can infer that earlier GA at delivery for twins is unlikely to explain twins’ persistently smaller size.
There are some limitations to using longitudinal data including loss-to-follow-up over time. However, linear mixed models are the favored approach to model growth trajectories when there is missing height or weight data.31 Nonetheless, only half of our analytic sample had measurements between ages 15–18y, meaning that we cannot rule out selection bias that may influence the later end of our trajectory; attrition was slightly higher for twins and subsequent missingness is unlikely to be completely at random. Loss of statistical significance at the end of the age distribution for height may reflect a loss of statistical power. We were also unable to fully harmonize race and ethnicity between the cohorts, but the proportion of twins did not vary significantly between groups. We did not have information on height in MCS prior to age 3 but considering that height and BMI were both smaller at 3y in the combined cohort, it is unlikely that twins caught up to singletons between 0–3y. The chorionicity and zygosity of twins in this study was unknown, which may have implications for fetal growth. However, another longitudinal study found no difference in height by zygosity,9 while another among adults found that both monozygotic and dizygotic twins were lighter than singletons.21 Furthermore, our sample also likely included singletons with pathological growth restriction, meaning that it is unlikely that extremes of growth for either twins or singletons overly influenced our findings. However, our findings present a population-based average and may not generalize to the extremes of growth, such as the largest or smallest twins or singletons.
Supplementary Material
Evidence before this study
We searched Pubmed for research studies that evaluated height, weight, or BMI in twins. Twins are born smaller than singletons, but there is conflicting evidence on whether twins catch up to singletons on height, weight, or BMI in childhood, and risk of overweight or obesity in twins has not been explored. Quality of existing studies is mixed, with few studies that are longitudinal in nature or include appropriate singleton reference groups, limiting inference.
Added value of this study
Combining two large prospective cohorts with measurements through 18 years, we found that twins never caught up on height or weight to singletons, and may be proportionally smaller based on similar BMI at 18. Twins had lower risk of overweight/obesity throughout childhood but similar late adolescent risk.
Implications of all the available evidence
Twins being proportionally smaller than singletons may reflect early adaptive growth mechanisms that should be evaluated in the context of cardiometabolic risk. If smaller size is normal, twin-specific child growth charts may be warranted for use in identifying normal growth in pediatric practice.
Acknowledgments
Funding:
This research was supported by the Division of Population Health Research, Division of Intramural Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), National Institutes of Health. JLG, RS, and KLG have contributed to this work as part of their official duties as employees of the United States Federal Government.
Role of the Funder:
The funding organization was not directly involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Disclosures: The authors have no conflicts of interest to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Access to Data and Data Analysis:
Drs. Gleason and Sundaram had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Data Sharing Statement:
Deidentified data, as well as data dictionaries and details on informed consent are publicly available through the UK Centre for Longitudinal Studies and the US Bureau of Labor Statistics.
References
- 1.Grantz KL, Grewal J, Albert PS, et al. Dichorionic twin trajectories: the NICHD Fetal Growth Studies. Am J Obstet Gynecol 2016; 215(2): 221.e1-.e16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Muhlhausler BS, Hancock SN, Bloomfield FH, Harding R. Are twins growth restricted? Pediatr Res 2011; 70(2): 117–22. [DOI] [PubMed] [Google Scholar]
- 3.Hiersch L, Barrett J, Fox NS, Rebarber A, Kingdom J, Melamed N. Should twin-specific growth charts be used to assess fetal growth in twin pregnancies? Am J Obstet Gynecol 2022; 227(1): 10–28. [DOI] [PubMed] [Google Scholar]
- 4.McMillen IC, Robinson JS. Developmental origins of the metabolic syndrome: prediction, plasticity, and programming. Physiol Rev 2005; 85(2): 571–633. [DOI] [PubMed] [Google Scholar]
- 5.Zheng M, Lamb KE, Grimes C, et al. Rapid weight gain during infancy and subsequent adiposity: a systematic review and meta-analysis of evidence. Obes Rev 2018; 19(3): 321–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mericq V, Martinez-Aguayo A, Uauy R, Iniguez G, Van der Steen M, Hokken-Koelega A. Long-term metabolic risk among children born premature or small for gestational age. Nat Rev Endocrinol 2017; 13(1): 50–62. [DOI] [PubMed] [Google Scholar]
- 7.van Dommelen P, de Gunst M, van der Vaart A, van Buuren S, Boomsma D. Growth references for height, weight and body mass index of twins aged 0–2.5 years. Acta Paediatr 2008; 97(8): 1099–104. [DOI] [PubMed] [Google Scholar]
- 8.Ooki S, Yokoyama Y. Physical growth charts from birth to six years of age in Japanese twins. J Epidemiol 2004; 14(5): 151–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ali MA, Ashizawa K, Kato S, Kouchi M, Koyama C, Hoshi H. Biological variables in height growth of Japanese twins: a comparison with those of singletons. Ann Hum Biol 2007; 34(3): 283–95. [DOI] [PubMed] [Google Scholar]
- 10.Booman A, Foster BA, Lyon-Scott K, Marino M, Snowden JM, Boone-Heinonen J. Early childhood weight trajectory differences in twins, singletons, and gestational-age matched singletons. Prev Med Rep 2025; 50: 102984. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Eriksen W, Sundet JM, Tambs K. Adult body height of twins compared with that of singletons: a register-based birth cohort study of Norwegian males. Am J Epidemiol 2013; 177(9): 1015–9. [DOI] [PubMed] [Google Scholar]
- 12.Eriksen W, Tambs K. Is the twin-singleton difference in BMI related to the difference in birth weight? A register-based birth cohort study of Norwegian males. Am J Hum Biol 2016; 28(4): 566–73. [DOI] [PubMed] [Google Scholar]
- 13.Silventoinen K, Magnusson PK, Tynelius P, Kaprio J, Rasmussen F. Heritability of body size and muscle strength in young adulthood: a study of one million Swedish men. Genet Epidemiol 2008; 32(4): 341–9. [DOI] [PubMed] [Google Scholar]
- 14.Pietilainen KH, Kaprio J, Rissanen A, et al. Distribution and heritability of BMI in Finnish adolescents aged 16y and 17y: a study of 4884 twins and 2509 singletons. Int J Obes Relat Metab Disord 1999; 23(2): 107–15. [DOI] [PubMed] [Google Scholar]
- 15.Gleason JL, Yeung E, Sundaram R, et al. Longitudinal child growth patterns in twins and singletons in the Upstate KIDS cohort. Journal of Pediatrics 2023; In-press. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Buckler JM, Green M. A comparison of the early growth of twins and singletons. Ann Hum Biol 2004; 31(3): 311–32. [DOI] [PubMed] [Google Scholar]
- 17.Luke B, Leurgans S, Keith L, Keith D. The childhood growth of twin children. Acta Genet Med Gemellol (Roma) 1995; 44(3–4): 169–78. [DOI] [PubMed] [Google Scholar]
- 18.Buckler JM, Green M. The growth of twins between the ages of 2 and 9 years. Ann Hum Biol 2008; 35(1): 75–92. [DOI] [PubMed] [Google Scholar]
- 19.Estourgie-van Burk GF, Bartels M, Boomsma DI, Delemarre-van de Waal HA. Body size of twins compared with siblings and the general population: from birth to late adolescence. J Pediatr 2010; 156(4): 586–91. [DOI] [PubMed] [Google Scholar]
- 20.Estourgie-van Burk GF, Bartels M, van Beijsterveldt TC, Delemarre-van de Waal HA, Boomsma DI. Body size in five-year-old twins: heritability and comparison to singleton standards. Twin Res Hum Genet 2006; 9(5): 646–55. [DOI] [PubMed] [Google Scholar]
- 21.Andrew T, Hart DJ, Snieder H, de Lange M, Spector TD, MacGregor AJ. Are twins and singletons comparable? A study of disease-related and lifestyle characteristics in adult women. Twin Res 2001; 4(6): 464–77. [DOI] [PubMed] [Google Scholar]
- 22.Plewis I, Calderwood L, Hawkes D, Hughes G, Joshi H. The millennium cohort study: technical report on sampling. London: Centre for Longitudinal Studies; 2007. [Google Scholar]
- 23.Connelly R, Platt L. Cohort Profile: UK Millennium Cohort Study (MCS). Int J Epidemiol 2014; 43(6): 1719–25. [DOI] [PubMed] [Google Scholar]
- 24.Grummer-Strawn LM, Reinold C, Krebs NF, Centers for Disease C, Prevention. Use of World Health Organization and CDC growth charts for children aged 0–59 months in the United States. MMWR Recomm Rep 2010; 59(RR-9): 1–15. [PubMed] [Google Scholar]
- 25.Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC Growth Charts for the United States: methods and development. Vital Health Stat 11 2002; (246): 1–190. [PubMed] [Google Scholar]
- 26.Baker PC, Mott FL. NLSY Child Handbook 1989: A Guide and Resource Document for the National Longitudinal Survey of Youth 1986 Child Data. Columbus Ohio: Center for Human Resource Research, The Ohio State University, 1989. [Google Scholar]
- 27.Cole TJ, Lobstein T. Extended international (IOTF) body mass index cut-offs for thinness, overweight and obesity. Pediatr Obes 2012; 7(4): 284–94. [DOI] [PubMed] [Google Scholar]
- 28.Rosenfield RL, Lipton RB, Drum ML. Thelarche, pubarche, and menarche attainment in children with normal and elevated body mass index. Pediatrics 2009; 123(1): 84–8. [DOI] [PubMed] [Google Scholar]
- 29.Bann D, Johnson W, Li L, Kuh D, Hardy R. Socioeconomic inequalities in childhood and adolescent body-mass index, weight, and height from 1953 to 2015: an analysis of four longitudinal, observational, British birth cohort studies. The Lancet Public Health 2018; 3(4): e194–e203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gleason JL, Lee W, Chen Z, et al. Fetal body composition in twins and singletons - are twins programmed to be smaller? JAMA Pediatrics 2025; 179(6): 630–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Schafer JL, Graham JW. Missing data: Our view of the state of the art. Psychol Methods 2002; 7(2): 147–77. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Drs. Gleason and Sundaram had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Deidentified data, as well as data dictionaries and details on informed consent are publicly available through the UK Centre for Longitudinal Studies and the US Bureau of Labor Statistics.