Young adult weight trajectories through midlife by body mass category

Authors

Errata

This article is corrected by:

  1. Errata: Erratum: Young adult weight trajectories through midlife by body mass category Volume 22, Issue 7, 1770, Article first published online: 13 May 2014

Correspondence: Eric A. Finkelstein (eric.finkelstein@duke-nus.edu.sg)

Abstract

Objective: To estimate the expected weight gain through midlife for those in a given BMI category in young adulthood.

Design and Methods: Group-based trajectory modeling and National Longitudinal Survey of Youth 1979 data from 1990 to 2008 were used to quantify weight trajectories through midlife for 10,038 young adult men and women stratified by BMI category. Logistic regression was used to assess the association of trajectory membership with obesity-related conditions (hypertension, diabetes, arthritis) in middle age.

Results: Annual weight gain averaged 0.53 kg (1.17 lb) across the entire sample. However, there was considerable variation by and within BMI categories. More than 98% of men and 92% of women were on upward-sloping trajectories, generally moving into a higher BMI category by middle age. Those who experienced early and rapid weight gain during young adulthood were most likely to be on a steeper trajectory and had greater risks for obesity-related conditions.

Conclusion: This study points to the health and weight benefits of entering young adulthood with a normal BMI, but further reveals that this is no guarantee of maintaining a healthy weight through midlife. For those who are young adults today, weight maintenance is unlikely to occur without significant environmental or technical innovation.

Introduction

BMI category is often used as a discriminating factor for eligibility for select obesity treatments. For example, the National Institutes of Health recommends bariatric surgery only for those with Class III obesity (BMI ≥ 40 kg/m2) or Class II obesity (BMI 35-39.9 kg/m2) combined with significant comorbidities [1]. The Food and Drug Administration recently extended the eligibility for laparoscopic adjustable gastric banding (LAGB) for one type of band to include adults with Class I obesity (BMI 30-34.9 kg/m2) [2]. Even for nonsurgical obesity interventions, eligibility or employer/insurer subsidies often hinge on meeting a defined weight class, such as a BMI of ≥30 kg/m2 [3].

Using BMI category as a selection criterion for select weight management initiatives may be appropriate if there is evidence that the interventions are neither safe nor effective for those who fall outside the BMI category. However, whether particular weight loss interventions are appropriate for a given BMI category depends not only on safety and efficacy, but also on weight and co-morbidity outcomes that can be expected to occur in the absence of the intervention(s). For example, extending bariatric surgery to those with Class I obesity may be appropriate if most are likely to move into a higher BMI category regardless, but it would be less appropriate if a large percentage is likely to maintain or even lose weight without the surgery.

Several studies have estimated average weight gain among adults in the general population or for specific population subsets [4]. For example, in a recent and highly cited study, Mozaffarian et al [5] reported average annual weight gain of 0.38 kg (0.84 lb) among nonobese adults in the US. However, they did not quantify weight gain or trajectories for obese adults. Sheehan et al quantified weight gain by race/gender strata. For each strata, they showed that weight gain was most rapid for those in their 20s, 30s, and 40s and then tended to level off and then decrease after age 60 [7]. However, neither they nor more recent growth curve or trajectory analyses by Clarke et al [4] or Østbye et al [6] provide separate estimates of likely weight trajectories by starting BMI category.

In this study, in efforts to provide clinicians and policy makers with information about population level, expected weight trajectories through midlife for young adults in various starting BMI categories, group-based trajectory modeling was used to quantify 18-year weight trajectories for men and women who start out as young adults who were either normal weight (BMI 18.5-24.9 kg/m2), overweight (BMI 25-29.9 kg/m2), Class I, Class II, or Class III obese. We then explore the extent to which being on a steeper trajectory increases the risk of select chronic diseases (hypertension, diabetes, arthritis) in middle age.

It is expected that both across and within each BMI category, there will be several weight trajectories, including some indicative of significant weight gain and others that remain stable or even slope down throughout the 18-year period. Quantifying the number and steepness of these trajectories, the percentage of the population that falls onto each, and potential early identification of those most likely to be on the steepest trajectories and their relative disease burden, will prove beneficial for those who may wish to use BMI category as a discriminating factor for access to select weight management initiatives.

Methods and Procedures

Data

The National Longitudinal Survey of Youth 1979 (NLSY79) is an ongoing nationally representative cohort of 12,686 Americans, aged 14 to 22 years at baseline (1979). Participants were followed up annually from 1979 to 1994 and biennially thereafter [8]. The current analysis was limited to 10,295 individuals with body weight reported in 1990, who were followed through 2008, the latest wave with available data. The decision to use the 1990 wave, when most participants were in their late 20s, as the threshold for defining the cohort was taken to ensure an adequate number of individuals in each BMI category and a reasonably long follow-up period, and to yield results that would be most relevant to the current population and environment. Over this 18-year period, weight was recorded 11 times (all waves except 1991). Height was recorded in five waves (1981, 1982, 1985, 2006, and 2008). Both weight and height are based on self-report.

BMI for each of the 10,295 participants was calculated for each wave in which weight was reported, using the first height value reported in the wave the participant was first aged ≥20 years. A small fraction (2.5%) of individuals who were underweight in 1990 (37 men and 220 women) were excluded from analysis because of concerns about unobserved illness and data quality. This left a final analysis sample of 10,038 (5,037 men; 5,001 women). For those with missing body weight data in any of the 1992 to 2006 waves, the average weight from the preceding and subsequent waves, if available, was used. For those with missing body weight in 2008, the weight from 2006, if available, was carried forward. If body weight was missing in two or more consecutive waves, no imputations were made and the data point was left as missing. The proportion of individuals with missing body weight data before imputation ranged from 17.0% (1992 wave for men) to 31.0% (2004 wave for men), and after imputation from 14.0% (1992 wave for men) to 28.1% (2004 and 2006 waves for men). On average, each individual contributed 8.76 weight values out of a maximum of 11, with most (66.7%) contributing all 11.

Statistical analyses

All analyses were stratified by gender. Group-based trajectory modeling (proc traj in SAS) [9] was used to delineate body weight trajectories. This approach uses maximum likelihood estimation techniques to fit the trajectory groups [9]. The first step was to identify the “average” (i.e., single group) body weight trajectory that best fit (on the basis of significance of polynomial terms) the longitudinal body weight data, stratified by gender, overall, and by BMI category in 1990. Then, to determine the appropriate number of trajectory groups and the functional form of the trajectory in each group within each BMI category in 1990, we considered several factors, including model fit statistics (% change in Bayesian Information Criterion [BIC]), significance of polynomial terms, value of group membership probability and of average posterior probability (entropy), and number of individuals classified in a particular group, leaning toward parsimony in number of trajectory groups.[6, 11]

To determine the appropriate number of trajectory groups in each BMI category in 1990, six models were fit, consecutively increasing the number of specified trajectory groups, separately for each gender/1990 BMI category. If the specification of an additional trajectory group resulted in less than 1% change in the BIC or less than 20 individuals in a group or less than 1% of the subsample in a group or a group with an average predicted posterior probability of membership (based on maximum probability assignment rule) value of less than 0.70, which indicates inadequate internal reliability, then the number of groups specified in the more parsimonious model (without the additional trajectory group) was chosen [12].

Within each model, the functional form of the trajectory for each group was based on the significance of the polynomial terms, starting with a quintic (up to time5) specification for each trajectory group, then iteratively dropping nonsignificant (p-value > 0.05) polynomial terms. Each model retained linear terms irrespective of significance [12].

Once an appropriate model was determined for a particular gender/1990 BMI category, the role of race/ethnicity (Black versus Hispanic and non-Hispanic/non-Black) in influencing the probability of being in a higher (relative to the lowest) trajectory was assessed by including race/ethnicity as a time-invariant risk factor in the model [10].

Although NLSY contains limited health data, a “40-and-over health module” was administered once a participant turned 40. We conducted a series of logistic regression analyses (adjusting for age, race, and educational status) to assess whether trajectory group membership predicted the likelihood of having each of three obesity-related conditions: hypertension, diabetes, and arthritis. The hypothesis was that even among individuals who were in the same BMI category in 1990, those belonging to the higher weight trajectories were more likely to have these conditions in middle age.

All analyses were conducted using SAS for Windows Version 9, Cary, North Carolina, USA, utilizing sampling weights to allow for generating nationally representative estimates [15]. IRB exemptions for the analyses were granted by Institutional Review Boards of Duke University Medical Center and the National University of Singapore.

Results

In 1990, the highest proportions of men and women in the sample were aged 28-30 years (men: 37.8%; women: 38.5%), were of “non-Black, non-Hispanic” race/ethnicity (men: 79.0%; women: 79.0%), had an educational level of high school or less (men: 56.7%; women: 54.4%), and a BMI in the normal range (men: 46.2%; women: 63.2%). Mean body weight for men was 82.5 ± 0.2 kg (181.5 ± 0.4 lbs) and for women was 66.3 ± 0.2 kg (144.4 ± 0.5 lbs) (Table 1).

Table 1. Baseline (1990) demographic and anthropometric characteristics of the study cohort, by gender
VariableMen n (weighted %a) N = 5037Women n (weighted %a) N = 5001
  1. kg: kilograms; SE: Standard error.
  2. aweighted by the sample weight for the 1990 wave.
  3. bn's and %'s do not add up to column N and 100%, respectively, as 18 (0.3%) men and 18 women (0.2%) had missing data.
Age, y  
25-271,536 (28.5)1,399 (28.3)
28-302,023 (37.8)1,994 (38.5)
31-331,478 (33.7)1,608 (33.2)
Mean age (y; weighted mean ± SE)29.2 ± 0.0329.2 ± 0.03
Race/ethnic group  
Non-Black, non-Hispanic2,848 (79.0)2,859 (79.0)
Black1,346 (14.3)1,303 (14.4)
Hispanic843 (6.7)839 (6.5)
Highest grade completedb  
Less than high school945 (12.7)749 (10.5)
High school2,199 (44.0)2,150 (43.9)
Some college (1-3 y)967 (20.3)1,160 (23.3)
4 or more years of college908 (22.8)924 (22.3)
Body mass index category  
Normal weight (18.5-24.9 kg/m2)2,258 (46.2)2,950 (63.2)
Overweight (25.0-29.9 kg/m2)2,083 (41.4)1,229 (22.9)
Obese I (30.0-34.9 kg/m2)541 (9.8)522 (9.0)
Obese II (35.0-39.9 kg/m2)109 (1.8)171 (2.8)
Obese III (≥40.0 kg/m2)46 (0.8)129 (2.2)
Mean weight (kg; weighted mean ± SE)
Overall82.5 ± 0.266.3 ± 0.2
Normal weight73.1 ± 0.258.4 ± 0.1
Overweight86.1 ± 0.272.2 ± 0.2
Obese I101.4 ± 0.485.8 ± 0.4
Obese II115.2 ± 1.298.3 ± 0.9
Obese III136.2 ± 2.5115.7 ± 1.2
Mean height (m, weighted mean ± SE)
Overall1.79 ± 0.0011.64 ± 0.001
Normal weight1.79 ± 0.0011.64 ± 0.001
Overweight1.79 ± 0.0011.63 ± 0.002
Obese I1.78 ± 0.0031.63 ± 0.003
Obese II1.76 ± 0.0091.63 ± 0.008
Obese III1.79 ± 0.0141.61 ± 0.006

The average 18-year weight trajectories defined for men and women overall (Figure 1) and for nearly all BMI categories (Figure 2) revealed gradual increases in body weight over time. The two exceptions were obese class III men, who showed a slight decrease in weight, and obese class II women, whose average trajectory fluctuated over the 18-year period. Intercept and polynomial terms defining these trajectories are presented in Table 2A.

Figure 1.

“Average” 18-year (1990–2008) weight trajectory for men and women

Figure 2.

“Average” 18-year (1990–2008) weight trajectory for A) men and B) women by body mass index category in 1990

Table 2. Intercept, linear, and polynomial terms (if any) defining the 18-year body weight trajectories for men and women. A. Intercept, linear and polynomial terms defining the “average” 18-year body weight trajectory for men and women, overall and stratified by body mass category in 1990
“Average” trajectory for…Intercept (SE)Linear term (SE)Quadratic term (SE)Cubic term (SE)Quartic term (SE)Quintic term (SE)
Men  
Overall82.56 (0.24)0.60 (0.08)0.05 (0.02)−0.006 (0.002)0.0002 (0.0001)
NW in 199073.13 (0.19)1.10 (0.13)−0.13 (0.05)0.018 (0.009)−0.0012 (0.0006)0.00003 (0.00001)
OW in 199086.16 (0.23)0.52 (0.11)0.06 (0.03)−0.007 (0.003)0.0002 (0.0001)
ObI in 1990101.02 (0.53)0.87 (0.12)−0.02 (0.01)
ObII in 1990114.09 (1.69)0.73 (0.20)
ObIII in 1990136.80 (3.86)−0.06 (0.23)
Women  
Overall66.35 (0.25)0.77 (0.13)−0.10 (0.06)0.021 (0.010)−0.0017 (0.0006)0.00004 (0.00001)
NW in 199058.40 (0.14)1.00 (0.12)−0.15 (0.05)0.022 (0.008)−0.0015 (0.0005)0.00004 (0.00001)
OW in 199072.15 (0.26)0.27 (0.21)0.15 (0.06)−0.014 (0.005)0.0004 (0.0001)
ObI in 199085.61 (0.56)0.89 (0.16)−0.03 (0.01)
ObII in 199098.62 (1.14)−2.21 (0.75)0.80 (0.20)−0.073 (0.019)−0.0020 (0.0006)
ObIII in 1990115.79 (1.56)0.17 (0.14)
B. Intercept, linear and polynomial terms (if any) defining the 18-year body weight trajectories in each 1990 body mass category for men
TrajectoryIntercept (SE)Linear term (SE)Quadratic term (SE)Cubic term (SE)Quartic term (SE)Quintic term (SE)
NW in 1990  
NW-161.54 (0.50)0.24 (0.03)
NW-268.18 (0.57)0.35 (0.03)
NW-373.12 (0.53)0.78 (0.08)−0.02 (0.004)
NW-477.87 (0.70)1.10 (0.06)−0.03 (0.003)
NW-583.06 (0.60)1.54 (0.11)−0.03 (0.01)
NW-686.19 (1.79)3.66 (0.37)−0.11 (0.02)
OW in 1990 
OW-175.73 (0.47)0.10 (0.07)0.08 (0.004)
OW-283.59 (0.67)0.39 (0.03)
OW-389.48 (0.44)0.60 (0.21)0.10 (0.05)−0.012 (0.005)0.0004 (0.0001)
OW-496.31 (1.48)1.36 (0.12)−0.02 (0.01)
OW-5101.61 (1.40)3.32 (0.43)−0.10 (0.03)
OBI in 1990 
ObI-188.69 (1.00)−0.68 (0.24)0.04 (0.01)
ObI-299.19 (1.11)0.29 (0.07)
ObI-3106.28 (0.73)1.39 (0.15)−0.04 (0.01)
ObI-4114.89 (2.43)1.80 (0.27)
OBII in 1990 
ObII-1103.77 (3.95)−0.32 (0.22)
ObII-2112.97 (2.10)0.72 (0.25)
ObII-3131.32 (2.41)1.38 (0.43)
OBIII in 1990 
ObIII-1136.80 (3.86)−0.06 (0.23)
C. Intercept, linear and polynomial terms (if any) defining the 18-year body weight trajectories in each 1990 body mass category for women
TrajectoryIntercept (SE)Linear term (SE)Quadratic term (SE)Cubic term (SE)Quartic term (SE)Quintic term (SE)
  1. NW, normal weight; OW, overweight; ObI, Obese Class I; ObII, Obese Class II; ObIII, Obese Class III; SE, standard error.
NW in 1990  
NW-151.99 (0.25)−0.01 (0.07)0.03 (0.01)−0.001 (0.0004)
NW-257.37 (0.30)0.73 (0.17)−0.13 (0.07)0.02 (0.01)−0.0013 (0.0007)0.00003 (0.00001)
NW-361.75 (0.40)1.12 (0.11)−0.05 (0.01)0.001 (0.0005)
NW-465.74 (0.74)1.64 (0.11)−0.04 (0.01)
NW-567.61 (1.17)3.70 (0.40)−0.11 (0.02)
OW in 1990
OW-166.24 (0.66)−3.42 (0.58)0.98 (0.19)−0.118 (0.029)0.0065 (0.0019)−0.0001 (0.00004)
OW-270.82 (0.82)0.06 (0.45)0.14 (0.08)−0.014 (0.007)0.0004 (0.0002)
OW-376.57 (0.77)1.49 (0.12)−0.04 (0.01)
OW-477.94 (1.43)4.19 (0.85)−0.15 (0.05)
OBI in 1990
ObI-176.84 (1.52)−0.93 (0.36)0.04 (0.02)
ObI-283.21 (1.41)0.17 (0.07)
ObI-388.88 (0.89)1.84 (0.28)−0.07 (0.02)
ObI-496.93 (2.37)4.03 (0.86)−0.15 (0.04)
OBII in 1990
ObII-195.41 (1.71)−2.89 (0.80)0.76 (0.22)−0.062 (0.019)0.0016 (0.0005)
ObII-2108.23 (3.72)3.10 (1.06)−0.13 (0.07)
OBIII in 1990
ObIII-1102.44 (2.82)−0.72 (0.34)
ObIII-2113.37 (2.68)0.17 (0.15)
ObIII-3129.13 (2.56)0.73 (0.24)

The predicted number of unique weight trajectories within each 1990 BMI category for men ranged from one (obese class III) to six (normal weight) (Figure 3). With the exception of the lightest weight group of obese class I (15.4%) and obese class II (28.5%) men, even with this finer stratification all of the trajectories sloped upward. The percentages on the steepest trajectories in each of the BMI categories ranged from 1.8% of normal weight men to 22.6% of obese class II men. Intercept and polynomial terms defining each of the trajectory groups for men are presented in Table 2B.

Figure 3.

18-year (1990–2008) weight trajectories for men, by 1990 body mass index category

Figure 4 presents the weight trajectories for women. The predicted number of trajectories ranged from two (obese class II) to five (normal weight). Unlike for men, with the exception of normal weight, there was at least one trajectory in each BMI category that sloped downward at some point. Although most women gained weight during this period, the proportion of women on a trajectory indicating weight loss over an extended period ranged from 15.6 % of obese class I women to 75.9% of obese class II women, although the latter group was heavier in 2008 than in 1990. Intercept and polynomial terms defining these trajectories are presented in Table 2C.

Figure 4.

18-year (1990–2008) weight trajectories for women, by 1990 body mass index category

Race/ethnicity did not influence trajectory group membership in any BMI category among men, and among obese class I, II, and III women. Among normal weight and overweight women, Blacks were more likely to be on one of the higher trajectories relative to the lowest trajectory (results not shown, available on request).

For many trajectories, the predicted BMI in 2008 was at least one BMI category higher than the starting category (Table 3). For example, the normal weight trajectory for men had a starting BMI of 22.8 in 1990 and a concluding BMI of 25.9, in the low end of the overweight range. The weighted proportion of men that moved up one BMI category was 52.3%; another 3.1% moved up two BMI categories. Only 0.5% moved down one BMI category. For women, 34.9% moved up one BMI category and 4.4% moved up two categories. Only 1.9% moved to a lower BMI category.

Table 3. Body mass index at start and end of 18-year weight trajectories, and associated annual weight change
Weight trajectory groupMenWomen
BMI (kg/m2) at …… of weight trajectory  BMI (kg/m2) at …… of weight trajectory  
startaendbAnnual weight change in kg (lbs)c%d, estartaendbAnnual weight change in kg (lbs)c%d, e
  1. BMI, body mass index; kg, kilograms; lbs, pounds; NW, normal weight; OW, overweight; ObI, obese Class I; ObII, obese Class II; ObIII, obese Class III.
  2. aPoint estimate of weight (kg) trajectory in 1990/Square of average height (m) for the participants in the trajectory.
  3. bPoint estimate of weight (kg) trajectory in 2008/Square of average height (m) for the participants in the trajectory.
  4. c(Point estimate of weight trajectory in 2008—Point estimate of weight trajectory in 1990)/18
  5. d%s for overall, NW in 1990, OW in 1990, ObI in 1990, ObII in 1990, and ObIII in 1990 are weighted by sample weight for the 1990 wave, while %s for trajectory groups within each body mass category are weighted by custom longitudinal weights derived from the NLSY website
  6. e%s for trajectory groups within each body mass index category add up to 100% to reflect the proportion in the entire population of men or women following that particular trajectory
Overall25.828.90.55 (1.20)100.024.828.30.52 (1.15)100.0
NW in 199022.825.90.56 (1.23)46.221.725.30.53 (1.18)63.2
NW-121.222.80.24 (0.54)4.720.421.90.20 (0.44)15.6
NW-222.124.10.35 (0.78)10.921.524.20.40 (0.87)22.0
NW-322.825.50.48 (1.05)13.722.426.50.63 (1.39)16.4
NW-423.326.80.64 (1.42)11.423.229.40.99 (2.17)7.6
NW-523.828.70.94 (2.07)4.723.333.81.69 (3.73)1.6
NW-624.833.61.71 (3.76)0.8----
OW in 199027.030.10.55 (1.21)41.427.130.80.55 (1.21)22.9
OW-126.127.70.25 (0.54)8.426.626.6−0.01 (−0.01)5.1
OW-226.628.80.39 (0.86)13.927.029.90.42 (0.93)8.6
OW-327.331.10.69 (1.53)12.027.432.30.76 (1.67)7.3
OW-427.932.70.92 (2.02)5.627.236.21.44 (3.18)1.8
OW-528.736.51.52 (3.35)1.5
ObI in 199031.835.00.57 (1.26)9.832.234.80.40 (0.87)9.0
ObI-131.431.60.03 (0.07)1.531.329.7−0.22 (−0.48)1.4
ObI-231.833.50.29 (0.65)3.932.033.10.17 (0.37)3.7
ObI-331.835.60.69 (1.53)3.632.136.30.64 (1.41)2.9
ObI-433.342.71.80 (3.96)0.833.241.41.33 (2.93)1.0
ObII in 199036.841.00.73 (1.62)1.837.338.80.23 (0.51)2.8
ObII-136.234.2−0.32 (−0.71)0.537.037.60.03 (0.07)2.1
ObII-236.740.90.72 (1.59)0.937.642.10.72 (1.60)0.7
ObII-337.844.91.38 (3.04)0.4
ObIII in 199042.742.40.06 (0.12)0.844.445.60.17 (0.38)2.2
ObIII-142.742.4−0.06 (−0.12)0.842.937.5−0.72 (−1.59)0.5
ObIII-243.144.30.17 (0.38)1.0
ObIII-347.352.10.73 (1.60)0.7

Although not all trajectories ended in a different BMI category in 2008 relative to 1990, as shown in Table 3, nearly all trajectories were associated with weight gain. On average, the cohort gained 0.53 kg (1.17 lbs) per year, with men showing average annual gains of 0.55 kg (1.20 lbs) and women showing average annual gains of 0.52 kg (1.15 lbs). As hypothesized, there was substantial variation around these estimates. Among men, the largest weight gain was observed among those with the steepest trajectories of the normal weight and obese class I categories, gaining 1.71 kg (3.76 lbs) and 1.80 kg (3.96 lbs) per year on average, respectively. For women, the largest weight gain occurred among those in the steepest trajectories of the normal weight (average annual gains of 1.69 kg [3.73 lbs]) and overweight (average annual gains of 1.44 kg [3.18 lbs]) categories. Only 1.3% of men and 7.4% of women were predicted to be on trajectories that were not associated with weight gain.

Table 4 presents results testing whether those on steeper trajectories within the same BMI category in 1990 were more likely to report having hypertension, diabetes, and arthritis in middle age. The results generally confirm this hypothesis. For most of the BMI categories for men and women, the weighted prevalence of these conditions was highest for those on the steepest trajectory. The odds of having these conditions was greater for those in the steepest as opposed to the flattest trajectory, and in most cases the odds ratios for those in the two middle categories fell somewhere in between.

Table 4. Prevalence and odds of hypertension, diabetes, and arthritis as self-reported in the 40 and over health module by weight trajectory group among men and women
MENWOMEN
Trajectory groupHypertensionbDiabetescArthritisdTrajectory groupHypertensionbDiabetescArthritisd
  1. aThe weighted proportion of men with missing data on health conditions ranged from 0% to 2.5% among those normal weight in 1990, ranged from 0% to 1.5% among those overweight in 1990, ranged from 0% to 0.9% among those obese Class I in 1990, was 0% among those obese Class II and obese Class III in 1990. The weighted proportion of women with missing data on health conditions ranged from 0% to 1.7% among those normal weight in 1990, ranged from 0% to 1.8% among those overweight in 1990, ranged from 0% to 2.5% among those obese Class I in 1990, ranged from 0% to 1.0% among those obese Class II in 1990, and ranged from 0% to 2.0% among those obese Class III in 1990.
  2. bBased on “Do you currently have hypertension or high blood pressure?”
  3. cBased on “Has a doctor ever told you that you have diabetes or high blood sugar?”
  4. dBased on “Have you ever had, or has a doctor ever told you that you have arthritis or rheumatism?”
  5. eNumbers represent weighted prevalence estimates, odds ratio (95% confidence interval). Weighted prevalence estimates are for the comorbid condition among individuals in the group, weighted by the sample weight for the wave in which the 40 and over health module was administered. Odds ratio reflects the adjusted (for age at wave of administration of 40 and over health module, race/ethnicity, and educational status) odds for the comorbid condition among individuals in that trajectory group relative to the flattest trajectory group.
Normal weight (NW) in 1990 a
NW-1 (N = 259)1.1%, 1.00 e1.0%, 1.0011.2%, 1.00NW-1 (N = 691)5.0%, 1.003.1%, 1.0014.3%, 1.00
NW-2 (N = 574)4.0%, 3.34 (3.27,3.42)1.5%, 1.40 (1.37,1.44)7.8%, 0.65 (0.64,0.66)NW-2 (N = 1089)3.8%, 0.75 (0.74,0.75)2.0%, 0.64 (0.63,0.65)10.3%, 0.69 (0.68,0.69)
NW-3 (N = 679)7.4%, 6.90 (6.75,7.05)1.8%, 1.71 (1.67,1.75)7.9%, 0.68 (0.68,0.69)NW-3 (N = 768)6.0%, 1.17 (1.16,1.18)3.6%, 1.19 (1.18,1.20)11.7%, 0.78 (0.78,0.79)
NW-4 (N = 499)6.9%, 6.40 (6.26,6.54)0.8%, 0.73 (0.71,0.75)9.4%, 0.82 (0.81,0.83)NW-4 (N = 325)11.6%, 2.36 (2.34,2.38)5.8%, 2.05 (2.03,2.07)14.8%, 1.05 (1.05,1.06)
NW-5 (N = 210)11.7%, 11.09 (10.86,11.34)3.7%, 3.86 (3.76,3.96)10.9%, 0.97 (0.96,0.98)NW-5 (N = 77)17.1%, 3.86 (3.81,3.90)3.2%, 1.08 (1.06,1.10)26.2%, 2.14 (2.12,2.16)
NW-6 (N = 37)21.8%, 22.62 (22.07,23.18)10.9%, 10.08 (9.80,10.37)12.5%, 1.10 (1.08,1.12)
Overweight (OW) in 1990 a
OW-1 (N = 483)5.9%%, 1.003.8%, 1.009.7%, 1.00OW-1 (N = 237)12.2%, 1.005.6%, 1.0012.1%, 1.00
OW-2 (N = 704)7.0%, 1.25 (1.24,1.26)3.5%, 0.95 (0.94,0.96)7.4%, 0.80 (0.80,0.81)OW-2 (N = 517)8.1%, 0.63 (0.63,0.64)3.2%, 0.55 (0.54,0.55)10.2%, 0.91 (0.91,0.92)
OW-3 (N = 579)12.5%, 2.48 (2.46,2.50)2.4%, 0.64 (0.64,0.65)10.4%, 1.17 (1.16,1.18)OW-3 (N = 387)17.5%, 1.51 (1.50,1.52)6.7%, 1.20 (1.19,1.21)16.5%, 1.55 (1.54,1.57)
OW-4 (N = 255)14.8%, 3.07 (3.05,3.10)2.7%, 0.74 (0.73,0.75)10.3%, 1.32 (1.31,1.34)OW-4 (N = 88)25.8%, 2.47 (2.45,2.50)4.2%, 0.71 (0.69,0.72)19.2%, 2.03 (2.01,2.05)
OW-5 (N = 62)17.4%, 3.61 (3.57,3.66)0.8%, 0.22 (0.21,0.23)16.7%, 2.18 (2.15,2.21)
Obese Class I (ObI) in 1990 a
ObI-1 (N = 91)15.6%, 1.0015.5%, 1.008.5%, 1.00ObI -1 (N = 70)21.7%, 1.0013.6%, 1.0024.7%, 1.00
ObI-2 (N = 218)14.1%, 1.06 (1.04,1.07)10.6%, 0.72 (0.71,0.73)10.4%, 1.34 (1.32,1.36)ObI -2 (N = 235)18.6%, 0.80 (0.79,0.81)12.7%, 0.91 (0.90,0.93)16.3%, 0.57 (0.56,0.58)
ObI-3 (N = 196)24.5%, 2.17 (2.15,2.20)14.5%, 1.03 (1.02,1.05)14.7%, 1.98 (1.94,2.01)ObI -3 (N = 171)11.2%, 0.42 (0.42,0.43)13.4%, 0.96 (0.95,0.98)27.5%, 1.14 (1.13,1.16)
ObI-4 (N = 36)37.6%, 4.03 (3.96,4.09)20.5%, 1.58 (1.55,1.60)25.6%, 4.05 (3.97,4.13)ObI -4 (N = 46)22.3%, 0.96 (0.95,0.98)14.3%, 1.03 (1.01,1.05)24.6%, 0.98 (0.96,0.99)
Obese Class II (ObII) in 1990 a
ObII-1 (N = 29)15.9%, 1.0010.8%, 1.008.5%, 1.00ObII -1 (N = 135)18.5%, 1.0013.5%, 1.0029.6%, 1.00
ObII-2 (N = 60)20.3%, 2.78 (2.71,2.86)8.8%, 1.28 (1.24,1.33)9.4%, 1.09 (1.05,1.12)ObII -2 (N = 36)38.0%, 3.09 (3.04,3.14)10.3%, 0.79 (0.77,0.81)40.5%, 2.11 (2.07,2.14)
ObII-3 (N = 20)22.6%, 4.01 (3.89,4.13)11.7%, 1.75 (1.69,1.81)18.9%, 2.43 (2.36,2.51)
Obese Class III (ObIII) in 1990 a
ObIII-1 (N = 46)25.5%, –26.8%, –27.4%, –ObIII -1 (N = 36)20.2%, 1.0022.9%, 1.0012.8%, 1.00
ObIII -2 (N = 55)35.5%, 2.32 (2.27,2.37)22.2%, 0.96 (0.94,0.98)24.7%, 2.20 (2.15,2.25)
ObIII -3 (N = 38)29.8%, 2.10 (2.06,2.15)30.6%, 1.57 (1.54,1.60)19.3%, 1.74 (1.69,1.78)

Discussion

Across all BMI categories, these results reveal that, on average, young adult men and women can expect to gain 0.55 kg (1.20 lb) and 0.52 kg (1.15 lb) per year, respectively. Across the 18 years of analysis, this translates into an average weight gain of between 9 and 10 kg or roughly 21 lbs, which represents an increase in BMI of 3.1 kg/m2 (or 12.0% relative to baseline) for men and 3.5 kg/m2 (or 14.1% relative to baseline) for women. With the exception of Obese III men, who managed to maintain their (high) weight, average single-group weight trajectories from young adulthood through midlife were nearly parallel across BMI categories for both men and women, as seen in Figure 2. This highlights the importance of maintaining a healthy weight throughout young adulthood. Although there is likely to be weight gain over the next 18 years, weight at middle age and obesity-related conditions are likely significantly lower for those who enter young adulthood with a lower BMI.

Although informative, the average weight trajectories mask the variation that occurs within each BMI category. Yet, although this variation exists, in few cases were we able to identify trajectories that sloped down. Only 0.5% of men and 1.9% of women were on a trajectory that resulted in more than a one unit decrease in BMI over the 18-year period. This result is striking as it points to the reality that few young adults are able to maintain their weight through middle age. Given the rapidly increasing occurrence in obesity seen over the past several decades, a significant fraction of the weight gain observed in this cohort is likely caused by environmental factors, such as increased use of technology and increased availability of calorie dense foods. The heterogeneity of weight gain within BMI categories should also be considered when designing or evaluating weight loss interventions.

As noted in the Introduction, bariatric surgery is increasingly being considered for obese individuals, including those with Class I obesity [2]. For young adult men and women in this BMI category in 1990, roughly 44% gain enough weight to meet the (Class II Obesity) BMI eligibility under the former LAGB guidelines [16]. The trajectory analysis further reveals that those who experience significant weight gain in their early 30s are most likely to be on the steepest trajectories, thus resulting in long-term significant weight gain. As such, they may be the most appropriate candidates for obesity surgery or other aggressive weight management interventions.

The results also reveal that 24% of Obese III women were on a trajectory that resulted in weight loss that reduced their BMI from 42.9 kg/m2 to 37.5 kg/m2. This reduction was enough to reduce their risk of the selected obesity-related conditions relative to those who started in the same BMI class in 1990 but who did not lose weight, although their risks remained higher than risks for most of those who started in lower BMI classes. For Obese III men and Obese II men and women, the results provide little evidence that behavioral weight management efforts that were available during this time period were effective in generating sustained weight loss for this population subset.

Finally, although most weight management programs target those who are obese, these results revealed that 46% of overweight men and 40% of overweight women were on trajectories that led to obesity and increased disease risk by middle age, thus revealing that weight management initiatives should also target young adults in the overweight range. A further 66% of normal weight men and 41% of normal weight women were on a trajectory that led to either overweight or obesity. In each case, it was those who showed higher than average weight gain in their early 30s who were predicted to experience the largest weight gain. In fact, this result holds for all BMI classes. Thus, early and rapid weight gain during young adulthood can be used as a marker for who might be most appropriate for more aggressive weight management interventions across all BMI categories.

This analysis is subject to several limitations. First, BMI was based on self-report. This is likely to underestimate actual weight, and more so for those in higher BMI categories as there is differential underreporting of weight with increasing BMI [17]. Missing data also increases in later waves, where people are likely at a higher weight. Therefore, if actual weights were available for all individuals in all waves, the trajectories would likely shift upward slightly and be even steeper, thus making the results conservative. Second, although the analysis method is generally robust to missing data [14], if the data are not missing at random then the trajectory estimates would be biased. However, the extent to which missing BMI data may have biased the slope of the trajectories is not clear.

NLSY79 gathered no information on diet or physical activity of participants prior to 2000. Thus, we were unable to utilize these variables either as predictors of trajectory group membership (available prior to trajectory onset) or as time-varying covariates, which might be available at various time points throughout the trajectory period [11]. Further, the survey did not gather information on the involvement of its participants in weight loss efforts or lifestyle interventions. Therefore, it was not possible to test the extent to which participation in specific weight management programs influenced weight trajectories. However, we can say that on a population-wide level, whatever efforts were taking place were largely ineffective at generating sustained weight loss, although it is possible that weight gain might have been even larger in their absence. Because of sample size considerations, race-specific trajectories by gender and BMI strata were not examined. However, analyses did assess whether being Black influenced the probability of being on a higher (relative to the lowest) trajectory group. The findings suggest that racial disparities in weight gain exist only among women who are normal weight or overweight in young adulthood; weight gain patterns among those who are already obese by their late 20s appear similar across Whites and Blacks. No racial disparities in weight gain patterns among men were found.

In conclusion, the increasing prevalence of obesity has been well established. Less well-known is the stark reality that few adults in their late 20s to early 30s are able to maintain their weight throughout adulthood; we show that roughly 95% of young adults, regardless of weight loss efforts, gain weight as they age. Though some weight gain over time because of aging and slowing metabolism is to be expected, in many cases the weight gain is significant and greatly increases risk for chronic disease. This analysis points to the weight and health benefits of entering young adulthood with a normal BMI, but further reveals that this is no guarantee of maintaining a healthy weight in middle age. Although the analysis identified that those experiencing rapid weight gain in young adulthood are most likely to benefit from aggressive weight management efforts, it further highlights that, for the vast majority of today's young adults, weight maintenance is unlikely to occur without significant environmental or technical innovation.

Conflicts of interest: Dr. Finkelstein is a consultant for Allergan, Inc., and has formerly consulted with Ethicon Endosurgery. Neither of these companies have any involvement in the current analyses. Dr. Østbye has formerly consulted with Eli Lily, Astra, and Utah State University. None of these companies or institutions has any involvement in the current analyses. The remaining authors have no conflicts of interest.

Acknowledgments

The analyses were in part funded by an A*STAR infrastructure grant to the Duke-NUS Program in Health Services and Systems Research. RM, TO, and EF conceived the analysis. RM analysed the data. RM, TO, CR, and EF interpreted the results of the data analysis. All authors were involved in writing the paper and had final approval of the submitted and published versions.

Ancillary