Skip to main content
NCBI home page
The Journals of Gerontology Series B: Psychological Sciences and Social Sciences logoLink to The Journals of Gerontology Series B: Psychological Sciences and Social Sciences
. 2019 Mar 7;75(8):1689–1698. doi: 10.1093/geronb/gbz022

Evidence of Bidirectional Associations Between Depressive Symptoms and Body Mass Among Older Adults

Charles F Emery 1,✉, Deborah Finkel 2,3, Margaret Gatz 4,5, Anna K Dahl Aslan 3,5
Editor: Lynn M Martire
PMCID: PMC7489100  PMID: 30843043

Abstract

Objectives

Body fat, measured with body mass index (BMI), and obesity are associated with depressive symptoms. Among younger adults there is stronger evidence of obesity leading to depressive symptoms than of depressive symptoms leading to obesity, but the temporal relationship is unknown among older adults. This study utilized dual-change-score models (DCSMs) to determine the directional relationship between body mass and depressive symptoms among older adults.

Method

Participants (n = 1,743) from the Swedish Twin Registry (baseline age range 50–96 years) completed at least one assessment of BMI (nurse measurement of height and weight) and the Center for Epidemiologic Studies-Depression scale (CESD). More than half the sample completed 3 or more assessments, scheduled at intervals of 2–4 years. DCSMs modeled the relationship of BMI and CESD across age, both independently and as part of bivariate relationships.

Results

Depressive symptoms contributed to subsequent changes in BMI after age 70, while BMI contributed to subsequent changes in depressive symptoms after age 82. Thus, there is a reciprocal relationship that may change with age. The effect was more pronounced for women.

Discussion

The association of BMI and depressive symptoms is bidirectional among older adults, and it appears to be affected by both age and sex.

Keywords: Body mass index, Depression, Dual change score models, Women’s health

Obesity (typically defined by a body mass index [BMI] ≥ 30) and overweight (25 ≤ BMI < 30) continue to present major public health concerns due to associated serious medical comorbidities, such as hypertension, diabetes, and heart disease (Lavie, Milani, & Ventura, 2009), as well as psychological distress (e.g., greater symptoms of depression and anxiety; Goldney & Wittert, 2009). The interrelationship of body fat with symptoms of psychological distress is especially of interest due to their documented association. A meta-analysis of cross-sectional studies revealed that depression was associated with an 18% increased risk of being obese (de Wit et al., 2010), and data suggest that adults with extreme (Class III) obesity (BMI ≥ 40; Onyike, Crum, Lee, Lyketsos, & Eaton, 2003) may be at particularly high risk of developing depression. Interestingly, the latter result was observed in women but not among men.

Most past studies of body fat and depression are cross-sectional, but there have been a number of well-designed longitudinal studies, 15 of which were included in a meta-analysis designed to explore the question of causality, that is, whether depression is predictive of overweight/obesity or overweight/obesity status is predictive of depression (Luppino et al., 2010). Results indicate a bidirectional relationship, with overweight/obese individuals experiencing a significantly increased risk of developing depression over time (55% increase for obese, 27% increase for overweight); and depressed individuals having a 58% increased risk of becoming obese, but no significant increase in risk of becoming overweight. A subsequent systematic review of 22 longitudinal studies evaluating the relationship of obesity to depression provides stronger evidence in support of the direction from obesity to depression than from depression to obesity (Faith et al., 2011). Although cross-sectional studies indicate that the relationship between depression and obesity is greater among women than men, longitudinal studies are equivocal, with meta-analytic data indicating that sex is not a moderator (Luppino et al., 2010) while data from the systematic review suggests that it is a moderator (Faith et al., 2011). Several mechanisms have been postulated to explain the interrelationship of BMI and depression. Obesity is thought to lead to depression via negative self-image or impaired physical function (Luppino et al., 2010), while depression may lead to weight gain and obesity via unhealthy behaviors (e.g., high intake of calorie dense food, low exercise activity), reduced sleep quality, and side effects of anti-depressant medication (Faith et al., 2011).

Obesity and Depression Among Older Adults

Most studies in this area have not been conducted with older adults, therefore the results of meta-analyses and systematic reviews reflect data primarily from young and middle-aged adults. The few studies conducted among older adults (or with a large proportion of older adults in the sample) provide inconsistent results. Earlier cross-sectional studies with older adults indicate that being overweight is associated with depression, via the negative effects of dieting and poorer physical health (Ross, 1994); but that body weight is inversely associated with depression in older men and unrelated to depression in older women (Palinkas, Wingard, & Barrett-Connor, 1996). Epidemiologic data from a United States sample of adults up to age 96 reveal that obesity is highly correlated with major depression (Ohayon, 2007). However, studies of older adults in China indicate that higher BMI is associated with lower rather than higher depressive symptoms (Ho, Niti, Kua, & Ng, 2008). The authors interpret the latter results as evidence that older adults with higher muscle mass (consistent with greater BMI) are likely to have more functional reserve which may contribute to lower depressive symptoms. Longitudinal studies among older adults document an association of greater BMI with elevated depressive symptoms at 3-year follow-up (Sachs-Ericsson et al., 2007); and among adults over age 50, obesity is associated with increased risk of depression at 5-year follow up, but depression is not associated with elevated risk of obesity at follow-up (Roberts, Deleger, Strawbridge, & Kaplan, 2003).

Data from the more recent cross-sectional and longitudinal studies provide relatively strong support for a relationship between obesity and depression among older adults. Because subclinical levels of both BMI (e.g., overweight but not obese) and depression (e.g., depressive symptoms that do not meet a clinical threshold) have been linked to negative health outcomes (Field et al., 2001; Kubzansky, Davidson, & Rozanski, 2005), it is important to evaluate the degree to which they are related, and the extent to which one variable may contribute to change in the other. BMI is particularly relevant among older adults because higher BMI is a risk for morbidity and mortality during the earlier years of older adulthood (Peeters et al., 2003), but over age 80 there is evidence that BMI may become less closely associated with mortality (Dahl et al., 2013) and that higher BMI, paradoxically, is associated with survival, especially among individuals with chronic illness (Curtis et al., 2005; Weiss et al., 2008). Thus, it is important to examine the interrelationship of BMI with depression across a wide age range and to determine the degree to which the relationship may change over time.

Predicting Change in BMI and Depression Over Time

The optimal approach to clarify the directional relationship of BMI and depressive symptoms among older adults is longitudinal modeling of repeated evaluations among individuals who exhibit a range of levels of body fat and depressive symptoms. To address the question of cause and effect, it is necessary to evaluate the extent to which each component (BMI and depressive symptoms) predicts subsequent changes in the other component over time using structural equation models that allow for dynamic interaction between the two components. The development of dual-change-score models (DCSMs) to characterize age changes has facilitated specification and testing of dynamic age-related hypotheses (McArdle, 2001; McArdle, Hamagami, Meredith, & Bradway, 2000). These models assist with measuring the extent to which one variable influences subsequent changes in a second, related variable. The analyses for this study examined temporal dynamics between estimated changes in body mass and estimated changes in depressive symptoms among older adults ranging in age from 50 to 96 at baseline with the goal of (a) evaluating the reciprocal influences of BMI and depression across age; (b) identifying directional changes in the relationship; and (c) exploring sex differences in the interrelationship of BMI and depressive symptoms. Based on prior research in younger samples, it was hypothesized that (a) BMI and depression would exhibit reciprocal effects, (b) the directional relationship from BMI to depression would be stronger than the relationship from depression to BMI, and (c) any effects would be more pronounced among women than among men.

Method

Participants

Data came from three studies: Swedish Adoption/Twin Study of Aging (SATSA; Finkel & Pedersen, 2004), Origins of Variance in the Oldest Old (OCTO-Twin; McClearn et al., 1997), and Aging in Women and Men: A Longitudinal Study of Gender Differences in Health Behavior and Health among the Elderly (GENDER; Gold, Malmberg, McClearn, Pedersen, & Berg, 2002). Accrual procedures for the three studies have been described previously. In brief, all three samples were drawn independently from the population-based Swedish Twin Registry (Lichtenstein et al., 2002), resulting in nonoverlapping samples that were evaluated utilizing similar protocols. In-person testing (IPT) took place in a location convenient to participants, such as district nurses’ offices, health care schools, long-term care clinics, or at the participant’s home. Self-report questionnaires (including the measure of depressive symptoms used in this study) were mailed to participants before the scheduled IPT, and participants submitted the completed questionnaires at the time of the IPT. IPT intervals across the studies ranged from 2 to 4 years. Both BMI and depressive symptom data were available from at least one testing occasion for 1,743 individuals: 749 from SATSA, 532 from OCTO-Twin, and 462 from GENDER. Three or more waves of data were available for 55% of the sample. As shown in Table 1, the study sample was 59% female, with a mean age (at intake) of 74 (±11) years (range: 50–96 years). Due to sampling and procedural variations across the three studies, there were sample differences in proportion of women, age range and mean age at intake, and BMI at intake, as shown in Table 1. In addition, the mean number of IPTs and the mean interval between IPTs differed across the three studies. Also, mean scores for depressive symptoms were lower in the OCTO-Twin sample than in the other two samples.

Table 1.

Sample Demographics and Baseline Data

Variable SATSA OCTO-Twin GENDER Full sample
N individuals 749 532 462 1,743
% Female 59.6%a 66.4%b 49.8%c 59.0%
Mean age at intake (SD)* 65.82 (8.91)a 83.15 (2.74)b 74.42 (2.60)c 74.09 (10.96)
Age range at intake* 50–96 79–93 70–81 50–96
Mean BMI at intake (SD) 25.55 (3.99)a 24.58 (3.71)b 26.53 (3.84)c 25.09 (3.89)
BMI range at intake 16.44–46.07 13.84–36.45 16.91–39.32 13.84–46.07
Mean CESD at intake (SD) 10.64 (8.22)a 8.08 (8.15)b 10.26 (7.50)a 9.51 (8.28)
CESD range at intake 0–43 0–41 0–37 0–43
Max no. of waves 7 5 3 7
Mean no. of waves (SD) 4.18 (2.13)a 3.56 (1.42)b 2.44 (0.82)c 3.53 (1.80)
Mean wave interval (SD)* 3.45 (1.83)a 2.01 (0.04)b 3.97 (0.34)c 2.84 (0.83)
Open in a new tab

Note. BMI = body mass index; CESD = Center for Epidemiological Studies Depression Inventory.

*Data in years; values within each row with different superscripts indicate statistically significant differences, p < .01.

Among the three studies, the shortest interval between testing waves was 2 years. Therefore, to support statistical modeling and to maximize the age range available for inclusion in the study, data were divided into 23 two-year age intervals from 50 to 96 years: for example, everyone with data (regardless of IPT wave) at ages 50–51.9 was included in the first age interval (number of data points = 57). Thus, each participant was included in as many different age intervals as possible, consistent with the number of waves of data available for the participant. Sample sizes for each variable in each age interval are provided in Supplementary Table 1. Sample sizes were largest in age intervals where the three studies overlap: for example, age interval 80–81.9 (N = 426) and age interval 82–83.9 (N = 490). The data were too sparse after age 96 to support statistical modeling; therefore, only data up to age 94–95.9 (N = 34) were included in these analyses.

Measures

Height (m) and weight (kg) were measured by a research nurse at each IPT for standard calculation of BMI {=weight (kg)/[height (m)]2}.

Depressive symptoms were measured at each assessment with the Center for Epidemiological Studies—Depression (CESD) inventory, a 20-item measure of frequency of depressive symptoms during the preceding week (Radloff, 1977). Each item is rated on a 4-point scale, and total scores range from 0 to 60, with scores ≥16 indicating clinically significant depressive symptoms.

Statistical Method

DCSMs were used to examine age changes in BMI and CESD both independently and as part of bivariate relationships. Extensive discussions of the model are available (McArdle, 2001; McArdle et al., 2004), as well as comparisons of DCSMs with latent growth curve models (Ghisletta & de Ribaupierre, 2005; Lövdén, Ghisletta, & Lindenberger, 2005). As presented in Figure 1, the model is based on latent difference scores that create a growth curve reflecting change from one age to another age (∆BMI and ∆CESD), rather than performance at a single age, which is modeled as a function of both constant change (αBMI and αCESD) that accumulates over time in an additive fashion as well as proportional change (βBMI and βCESD) based on the previous score. In the full DCSM model, αBMI and αCESD are set to 1 and the parameters βBMI and βCESD differ from zero to the extent that the longitudinal change is nonlinear. The bivariate DCSM allows for a coupling mechanism (γ) where change in BMI depends on the previous value of CESD, and vice versa.

Figure 1.

Figure 1.

Open in a new tab

Bivariate dual change score model. BMI and CESD (D) in each age interval (e.g., D50 and BMI50) are modeled. Error variances (σBMIu and σDu) are assumed to be constant at each age; αBMI and αD represent constant change and are related to the slope factors BMIs and Ds, respectively; βBMI and βD represent proportional change. The model includes an estimate for intercepts (BMI0 and D0), mean intercepts (μBMI0 and μD0), and mean slopes (μBMIs and μDs); asterisks indicate standardized versions of parameters; cross-trait coupling is indicated by γBMI and γD.

In the univariate model (i.e., the top half of Figure 1), it is possible to test a variety of hypotheses about the nature of changes with age. First, setting β to zero and comparing the model fit to the full model tests for nonlinear changes with age. Second, multiple values of β can be estimated, testing for different rates of change at different age intervals. Previous investigations of BMI and health outcomes among older adults suggested that BMI tends to decline after age 70, but that rates of decline may change again after age 80 or 82 (Dahl et al., 2013; Dahl, Reynolds, Fall, Magnusson, & Pedersen, 2014; Dey, Rothenberg, Sundh, Bosaeus, & Steen, 1999). Also, as shown in Table 1, the GENDER sample included the largest proportion of men, with ages at intake ranging from 70 to 81 years, further supporting examination of age intervals in evaluating the third hypothesis pertaining to sex differences in BMI and depression trajectories. Thus, models with three different β values were tested for ages 50–69.9, 70–81.9, and 82–95.9, then estimates of model parameters were compared across sex.

In the bivariate model, it is possible to evaluate dynamic hypotheses about temporal order of changes in variables through restrictions on model parameters. Four alternative models can be addressed. First, the relationship between the two variables may be bidirectional, such that BMI affects changes in CESD and CESD affects changes in BMI (i.e., both γBMI and γCESD are nonzero). In the most reduced model, no dynamic coupling among the variables is included in the model (i.e., γBMI = γCESD = 0). Next, a model can be tested in which the dynamic relationship functions in one direction only, either with BMI as a leading indicator of change in CESD (i.e., γCESD = 0), or with CESD as a leading indicator of change in BMI (i.e., γBMI = 0).

It is important to note that one of the fundamental assumptions of DCSM is that data are missing at random. Previous investigations of SATSA data suggest that participants who continue in the study are significantly different from those who drop out (e.g., personality ratings: Pedersen & Reynolds, 1998; cognitive ability: Dominicus, Palmgren, & Pedersen, 2006). Of most importance in age-based DCSM is demonstrating that the pattern of missing data does not differ by age. Mean number of waves of participation did not differ across age groups in OCTO-Twin and GENDER. Given the large age range, number of waves, and long follow-up period, participants in SATSA under age 65 generally participated in an average of one more testing wave than participants age 65 or older, due to loss of older participants to mortality.

Univariate and bivariate DCSM were fit to the data using Mplus (Muthén & Muthén, 2012–2014). Model fit was indicated by the log-likelihood (−2LL) and the root mean square error of approximation (RMSEA; Browne & Cudeck, 1993). Adequate fit of the model to the data is indicated when the RMSEA is less than or equal to 0.1 and an RMSEA of .05 or less indicates “close” fit. Hypotheses were tested by comparing model fit indices; nested models were compared using the likelihood ratio test obtained by taking the difference between the obtained model fits (−2LL) and testing its significance with the degrees of freedom equal to the difference in the number of parameters of the two models. The current analyses focused on individual performance by including a correction for relatedness of twin pairs in the modeling (i.e., twin pairness was coded and included in models). Furthermore, all models were run in two subsamples, composed of one member of each pair, to confirm consistency with full sample results.

Power for dynamic growth models was evaluated based on observed effect sizes between BMI and depression, accounting for pair status. The power to detect dynamic and coupling parameters was estimated at .82 for as few as three time-points for small effects (Cohen’s d = .2). Power increased to .94 when projecting to 7 time-points. Power to detect small to medium dynamic shifts in younger-old and older-old was calculated to range from .52 to .86, respectively (Cohen’s d = .25 –.56).

Results

As shown in Table 1, mean BMI of the sample at intake was at the low end of overweight (25.1), but BMI ranged from below normal (13.8) to extreme obesity (46.1). Mean CESD score at intake was below the cut-off for clinical depression (9.5), but scores ranged from no symptoms of depression to severe depressive symptoms (43).

Univariate Analyses

In the first step of the analysis, the univariate DCSM was fit separately to the two primary variables, BMI and CESD, to determine the shape of the change trajectory over age and to provide starting values for the bivariate DCSM model. Five models were fit to the data for each measure in the full sample, and results of model comparisons are presented in Table 2. First, as a baseline, the full model (Model 1) was fit to the data. Then in Model 2, the proportional change parameter β was set to zero to test for nonlinear change trajectories. Change in model fit for Model 2 was significant for both BMI and CESD, indicating nonlinear age changes for both variables. Models 3 through 5 tested different nonlinear trajectories by allowing β to vary systematically across the age range. Model 3 tested the possibility of an inflection point (change in trajectory) at age 70 by fitting two β parameters to the data: one estimating change up to age 69.9, and the second estimating change at age 70 and above. Model comparisons indicated significant improvement in model fit compared with the baseline model for both BMI and CESD. Model 4 tested an inflection point at age 82 instead of age 70. Again, model comparisons indicated a significant improvement in model fit for both variables. Model 5 incorporated both inflection points at ages 70 and 82 (i.e., three β parameters). This model also fit significantly better than Model 1 for both variables.

Table 2.

Univariate Model Comparisons

BMI CESD
Model No. of paramaters −2LL RMSEA −2LL RMSEA
Initial model testing
1. Base model 7 −13,156 .035 −17,878 .031
2. Set β = 0 6 −13,166a .035 −17,887a .032
3. β1 (50–69.9) and β2 (70–95.9) 8 −13,080 a .029 −17,862a .030
4. β1 (50–81.9) and β2 (82–95.9) 8 −13,134 a .033 −17,857a .030
5. β1 (50–69.9) and β2 (70–81.9) and β3 (82–95.9) 9 −13,078a .029 −17,851a .029
Model testing by sex
6. All parameters differ across sexes 18 −13,045 .025 −17,814 .027
7. Equate all parameters across sexes 9 −13,078b .026 −17,851b .028
8. Equate βs for women 16 −13,097b .028 −17,835b .028
9. Equate βs for men 16 −13,069b .026 −17,815 .027
Open in a new tab

aModel fit differs from fit of Model 1 at p < .01.

bModel fit differs from fit of Model 6 at p < .01.

Comparing Model 5 to Model 3 provides a test of improvement in model fit achieved by adding the inflection point at age 82 to the model that already included the inflection point at age 70. Likewise, comparing Model 5 to Model 4 provides a test for improvement in model fit achieved by adding the inflection point at age 70 to the model that already included the inflection point at age 82. For CESD, both of these model comparisons were significant (Model 5 vs Model 3: LRT = 11, df = 1, p < .01; Model 5 vs Model 4: LRT = 6, df = 1, p < .05) indicating that both inflection points were necessary for optimal fit of the model to the data. The resulting univariate change trajectory is the “no coupling” trajectory for CESD presented in Figure 2: inflection points at age 70 and 82 can be seen, although the change in slope at age 82 is more dramatic. For BMI, only the comparison of Model 4 to Model 3 achieved significance (LRT = 54, df = 1, p < .01), indicating that the inflection point at age 82 did not improve model fit over the model including the inflection point at age 70. For BMI and CESD, however, the RMSEA was minimized for the model incorporating both inflection points. The “no coupling” trajectory for BMI in Figure 2 clearly shows a change in slope for BMI at age 70, and a less distinct change in slope at age 82.

Figure 2.

Figure 2.

Open in a new tab

Change trajectories resulting from fitting the full coupling versus no coupling bivariate dual change score model to BMI and CESD in full sample (a), and separately for men (b) and women (c).

Additional univariate models were run to examine differences in models between sexes. Model 6 represented the full univariate model with separate parameters for men and women: this model included three β parameters and served as the baseline for subsequent model comparisons. In Model 7, all parameters were equated across sexes, resulting in a significant reduction in model fit for both BMI and CESD. In Model 8, the three β parameters were equated (reduced to one change trajectory with age) for women only. Model 9 tested the same reduction in β parameters and simplification in change trajectory for men. Model comparisons indicated that three β parameters (two inflection points) improved model fit for women for both BMI and CESD. In contrast, including three β parameters in the model improved fit for men only for BMI; a single nonlinear change parameter was sufficient to model change in CESD for men. In subsequent analyses, three β parameters were included for both variables in all versions of the subsequent bivariate models, to support cross-variable coupling models.

Bivariate Analyses

To examine temporal dynamics in the relationship between BMI and CESD, four models were compared: (a) full coupling between variables (where both γBMI and γCESD are estimated, (b) no coupling (both γ set to zero), (c) CESD as a leading indicator of change in BMI (set γBMI = 0), and (d) BMI as a leading indicator of change in CESD (γCESD = 0). The four models were tested in the full sample and then separately in men and women. As shown in Table 3, results of model comparisons demonstrate that none of the coupling parameters could be dropped without significantly reducing model fit. Therefore, model comparisons indicated bidirectional temporal dynamics between BMI and CESD in all cases. Note that models including three γ parameters (i.e., two inflection points in cross-variable dynamics) were also estimated, but the results did not differ from the trajectories reported here. Figure 2 presents the estimated bivariate trajectories for BMI and CESD, indicated as “full coupling,” for the full sample (a), and separately for men (b), and women (c). In the BMI figure, the coupling corresponds to CESD as the leading indicator; in the CESD figure, BMI as the leading indicator. The extent of coupling is indicated by divergence of the “full coupling” from the “no coupling” trajectories. Parameter estimates from the models are provided in Supplementary Table 2. For both men and women, the parameter estimates are greater for γBMI than for γCESD, indicating that the effect of BMI on CESD is greater than the effect of CESD on BMI.

Table 3.

Bivariate Model Comparisons

Full sample Men Women
Model No. of parameters −2LL RMSEA −2LL RMSEA −2LL RMSEA
1. Full coupling 25 −30,867 .024 −11,891 .037 −18,898 .030
2. No coupling 23 −30,914a .025 −11,914a .038 −18,923a .031
3. γBMI = 0 24 −31,003a .026 −11,913a .038 −18,924a .031
4. γCESD = 0 24 −30,922a .025 −11,902a .038 −18,913a .31
Open in a new tab

aModel fit differs from fit of Model 1 at p < .01.

Age trajectories presented in Figure 2 demonstrate that the longitudinal change trajectories for BMI are similar up to age 70, but BMI declines more slowly after age 70 when the impact of CESD is included in the model. Similarly, the longitudinal change trajectories for CESD are similar across models up to age 82, but then CESD demonstrates more gradual decreases with age when BMI is incorporated in the model. Thus, the impact of CESD on BMI occurs at the earlier inflection point (age 70), while the impact of BMI on CESD occurs at the later inflection point (age 82). The difference in this bivariate dynamic relationship between men and women also is evident in Figure 2b and c. Although model comparisons indicate that bivariate coupling of BMI and CESD is significant for both men and women, the effect is more pronounced for women than for men, as reflected by the trajectories among women being more separated than the trajectories among men.

Discussion

These results extend data from Luppino and colleagues (2010) suggesting a bidirectional relationship between depression and obesity, and are consistent with data from Faith and colleagues (2011) and Roberts and colleagues (2003), indicating that the magnitude of the effect of body mass on depressive symptoms appears to be greater than the magnitude of the effect of depressive symptoms on body mass. Further, depressive symptoms contribute to changes in BMI after age 70, while BMI contributes to changes in depressive symptoms after age 82. The effects appear to be stronger among women than among men. This bivariate interrelationship between BMI and depressive symptoms during older age is consistent with the notion of elevated inflammatory factors (e.g., interleukin 6) contributing to increases in both dimensions (Ambrosio et al., 2018). However, later decreases in both BMI and depressive symptoms may suggest the appearance of alternative or additional mechanisms. Inflammation is associated with reduced muscle mass and lower muscle strength in well-functioning adults over age 70 (Visser et al., 2002). Thus, the inflammatory mechanism may help explain both earlier increases in BMI as well as later decreases in BMI. Declines in depressive symptoms may reflect additional psychosocial processes (e.g., lower emotional responsiveness, better adaptation to stressful life events; Jorm, 2000) that are less closely associated with inflammatory processes.

Trajectories of change in BMI and depressive symptoms

Univariate results indicate that the trajectory of change with age for BMI differs from that of depressive symptoms. BMI and depressive symptoms both appear to increase up to approximately age 70, after which BMI begins a gradual decline, while depressive symptoms level off until about age 82 before beginning to decline. The pattern of change in BMI is generally consistent with prior research (Dahl et al., 2014), and the pattern of changes in depressive symptoms is consistent with a review of cross-sectional studies (Jorm, 2000), but contrasts with longitudinal data indicating modest increases in depressive symptoms among older adults (e.g., Fiske, Gatz, & Pedersen, 2003). Because the OCTO-Twin participants had lower average depression scores (as shown in Table 1) and were more highly represented in the available data above age 82, all models were run a second time with only SATSA participants, providing a subsample covering the entire study age range. Results of these analyses revealed the same pattern. Thus, the gradual reduction in depressive symptoms starting after age 82 does not appear to be driven by the inclusion of OCTO-Twin participants at that age.

These results may help explain the interrelationship of body fat and depression among older adults. Although body fat and depressive symptoms appear to be positively correlated, above age 70 the leveling of BMI and gradual decrease may be attenuated by decreases in depressive symptoms. Thus, decreasing symptoms of distress may contribute to a less steep decline in BMI. Alternately, when BMI is included in the model of change in depressive symptoms, the decline in depressive symptoms is less steep, suggesting that decreasing body mass among older adults may contribute to increasing depressive symptoms. This bidirectional relationship between BMI and depressive symptoms among older adults may appear counterintuitive because lower levels of each variable contribute to increases in the other. However, it is possible that the bidirectional relationship is influenced by a third factor such as chronic illness, which might contribute to changes such as lower body fat and higher depressive symptoms. In addition, the physical health correlates of elevated BMI may vary with age. Although BMI elevations at any adult age are associated with chronic health conditions such as diabetes, among older adults a BMI in the overweight range (25 ≤ BMI < 30) may be protective of health and, in turn, associated with survival (Weiss et al., 2008). Thus, the negative relationship between higher BMI and lower depressive symptoms observed in this study at older ages may reflect enhanced well-being among those with greater physical reserve.

Sex Differences

Results of this study extend the review data from Faith and colleagues (2011) to older adults, suggesting that the relationship between body mass and depressive symptoms is more evident among women than among men. However, the greater magnitude of changes among women may contribute to larger bivariate effects. Indeed, older women exhibited greater decreases in body mass and depressive symptoms than older men. Depressive symptoms remained relatively flat among older men across age groups, and the decline in body mass after age 70 was attenuated among men. These data are consistent with prior studies indicating that women report greater depressive symptoms than men in middle-age and well into older adulthood, but that depressive symptoms then may be lower in older women than in older men (Barefoot, Mortensen, Helms, Avlund, & Schroll, 2001).

Strengths and Weaknesses of the Study

Strengths of this study include the use of longitudinal data with up to 19-year follow-up in one subsample (SATSA), and data modeled over a 46-year age span (50–96 years) across the three subsamples. In addition, the study had a relatively large sample, and utilized objectively measured height and weight for calculation of BMI. Many studies, especially larger community or epidemiological studies, have relied on self-report of weight which may underestimate (or overestimate) body fat calculations, especially among older adults. The study also included large proportions of both men and women in the sample. A relative weakness of the sample is the absence of longitudinal data for the full sample over the span of years from age 50 to 96. Although the modeling approach facilitates estimation of trajectories of BMI and depressive symptoms, the data do not reflect 46-year changes in BMI or depressive symptoms for any participants. Model parameters are resolved at the within-person level, but DCSMs estimate one set of between-person parameters that apply to the entire sample. Also, the statistical modeling did not include covariates, such as presence of chronic illness or education level, that might influence the relationship of BMI and depressive symptoms. In addition, participants in this study were primarily white, and recent data suggest that the association of BMI and depressive symptoms is significant for white women, but not for black women or for men (Carter & Assari, 2017). Thus, generalizability of these results to nonwhite individuals is unknown.

Conclusions

Overall, the data suggest that aging may be associated with increases in both body mass and depressive symptoms into old age (~70 years), and that both body mass and depressive symptoms then decline, on average, but rates of decline may differ in the two variables across men and women. Recent twin research indicates that the association of BMI and depressive symptoms is mainly mediated by genetic factors (Jokela et al., 2016), but further evaluation is needed of mechanisms, including effects of behavioral factors such as physical exercise and diet as well as biological factors such as inflammation and chronic disease, to help explain the interrelationship between changes in body mass and depressive symptoms among older adults. These data also suggest that mechanisms linking BMI and depressive symptoms may change with age. For example, behavioral mechanisms may be more relevant in one stage of old age, while biological mechanisms may be relevant at a different stage. Alternatively, biological mechanisms may be more relevant in the context of more extreme levels of BMI, while behavioral mechanisms influence the relationship at moderate levels of BMI. Questions pertaining to biological and behavioral mechanisms will be important to address in future studies. Also, with depressive symptoms and body mass more closely associated in women than men, sex-related influences (e.g., hormones, psychosocial differences) will be important to examine. In the meantime, future research should consider the interrelationship of depressive symptoms and BMI among older adults who may not exhibit clinical elevations on either dimension, paying particular attention to the relative increases in depressive symptoms among older adults, especially women, that appear to precede changes in BMI.

Funding

The Swedish Adoption/Twin Study of Aging (SATSA) was supported by National Institute on Aging (grants R01 AG04563 and R01 AG10175), the MacArthur Foundation Research Network on Successful Aging, the Swedish Council For Working Life and Social Research (FAS; 97:0147:1B, 2009-0795), and Swedish Research Council (825-2007-7460 and 825-2009-6141). Origins of Variance in the Oldest Old (OCTO-Twin) was supported by grant R01 AG08861. Aging in Women and Men: A Longitudinal Study of Gender Differences in Health Behavior and Health among the Elderly (GENDER) was supported by the MacArthur Foundation Research Network on Successful Aging, The Axel and Margaret Ax:son Johnson’s Foundation, The Swedish Council for Social Research, and the Swedish Foundation for Health Care Sciences and Allergy Research. This work was also supported by National Institute on Aging (grants 1R03AG048850-01 and 2R56AG037985-06).

Supplementary Material

gbz022_suppl_Supplemental-Table-1
Click here for additional data file. (13KB, docx)
gbz022_suppl_Supplemental-Table-2
Click here for additional data file. (12.9KB, docx)

Acknowledgments

The authors wish to thank Dr Nancy Pedersen and Dr Chandra Reynolds for early input into the conceptual framework of this investigation, and for providing comments on a draft of the article.

Conflict of Interest

None reported.

References

  1. Ambrosio G., Kaufmann F. N., Manosso L., Platt N., Ghisleni G., Rodrigues A. L. S.,…Kaster M. P (2018). Depression and peripheral inflammatory profile of patients with obesity. Psychoneuroendocrinology, 91, 132–141. doi:10.1016/j.psyneuen.2018.03.005 [DOI] [PubMed] [Google Scholar]
  2. Barefoot J. C., Mortensen E. L., Helms M. J., Avlund K., & Schroll M (2001). A longitudinal study of gender differences in depressive symptoms from age 50 to 80. Psychology and Aging, 16, 342–345. doi:10.1037/0882-7974.16.2.342 [DOI] [PubMed] [Google Scholar]
  3. Browne M., & Cudeck R (1993). Alternative ways of assessing model fit. In Bollen K., & Long S. (Eds.), Testing structural equation models (pp. 136–162). Beverly Hills, CA: Sage. [Google Scholar]
  4. Carter J. D., & Assari S (2017). Sustained obesity and depressive symptoms over 6 years: Race by gender differences in the Health and Retirement Study. Frontiers in Aging Neuroscience, 8, 312. doi:10.3389/fnagi.2016.00312 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Curtis J. P., Selter J. G., Wang Y., Rathore S. S., Jovin I. S., Jadbabaie F.,…Krumholz H. M (2005). The obesity paradox: Body mass index and outcomes in patients with heart failure. Archives of Internal Medicine, 165, 55–61. doi:10.1001/archinte.165.1.55 [DOI] [PubMed] [Google Scholar]
  6. Dahl A. K., Fauth E. B., Ernsth-Bravell M., Hassing L. B., Ram N., & Gerstof D (2013). Body mass index, change in body mass index, and survival in old and very old persons. Journal of the American Geriatrics Society, 61, 512–518. doi:10.1111/jgs.12158 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Dahl A. K., Reynolds C. A., Fall T., Magnusson P. K., & Pedersen N. L (2014). Multifactorial analysis of changes in body mass index across the adult life course: A study with 65 years of follow-up. International Journal of Obesity (2005), 38, 1133–1141. doi:10.1038/ijo.2013.204 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dey D. K., Rothenberg E., Sundh V., Bosaeus I., & Steen B (1999). Height and body weight in the elderly. I. A 25-year longitudinal study of a population aged 70 to 95 years. European Journal of Clinical Nutrition, 53, 905–914. doi:10.1038/sj.ejcn.1600852 [DOI] [PubMed] [Google Scholar]
  9. Dominicus A., Palmgren J., & Pedersen N. L (2006). Bias in variance components due to nonresponse in twin studies. Twin Research and Human Genetics: The Official Journal of the International Society for Twin Studies, 9, 185–193. doi:10.1375/183242706776382419 [DOI] [PubMed] [Google Scholar]
  10. Faith M. S., Butryn M., Wadden T. A., Fabricatore A., Nguyen A. M., & Heymsfield S. B (2011). Evidence for prospective associations among depression and obesity in population-based studies. Obesity Reviews: An Official Journal of the International Association for the Study of Obesity, 12, e438–e453. doi:10.1111/j.1467-789X.2010.00843.x [DOI] [PubMed] [Google Scholar]
  11. Field A. E., Coakley E. H., Must A., Spadano J. L., Laird N., Dietz W. H.,…Colditz G. A (2001). Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Archives of Internal Medicine, 161, 1581–1586. doi:10.1001/archinte.161.13.1581 [DOI] [PubMed] [Google Scholar]
  12. Finkel D., & Pedersen N. L (2004). Processing speed and longitudinal trajectories of change for cognitive abilities: The Swedish Adoption/Twin Study of Aging. Aging, Neuropsychology, and Cognition, 11, 325–345. doi:10.1080/13825580490511152 [Google Scholar]
  13. Fiske A., Gatz M., & Pedersen N. L (2003). Depressive symptoms and aging: The effects of illness and non-health-related events. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 58, 320–328. doi:10.1093/geronb/58.6.P320 [DOI] [PubMed] [Google Scholar]
  14. Ghisletta P., & de Ribaupierre A (2005). A dynamic investigation of cognitive dedifferentiation with control for retest: Evidence from the Swiss Interdisciplinary Longitudinal Study on the oldest old. Psychology and Aging, 20, 671–682. doi:10.1037/0882-7974.20.4.671 [DOI] [PubMed] [Google Scholar]
  15. Gold C. H., Malmberg B., McClearn G. E., Pedersen N. L., & Berg S (2002). Gender and health: A study of older unlike-sex twins. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 57, S168–S176. doi.10.1093/geronb/57.3.S168 [DOI] [PubMed] [Google Scholar]
  16. Goldney R. D., & Wittert G (2009). Obesity and depression or anxiety. British Medical Journal, 339, 871–872. doi:10.1136/bmj.b3868 [DOI] [PubMed] [Google Scholar]
  17. Ho R. C., Niti M., Kua E. H., & Ng T. P (2008). Body mass index, waist circumference, waist-hip ratio and depressive symptoms in Chinese elderly: A population-based study. International Journal of Geriatric Psychiatry, 23, 401–408. doi:10.1002/gps.1893 [DOI] [PubMed] [Google Scholar]
  18. Jokela M., Berg V., Silventoinen K., Batty G. D., Singh-Manoux A., Kaprio J.,…Kivimäki M (2016). Body mass index and depressive symptoms: Testing for adverse and protective associations in two twin cohort studies. Twin Research and Human Genetics: The Official Journal of the International Society for Twin Studies, 19, 306–311. doi:10.1017/thg.2016.14 [DOI] [PubMed] [Google Scholar]
  19. Jorm A. F. (2000). Does old age reduce the risk of anxiety and depression? A review of epidemiological studies across the adult life span. Psychological Medicine, 30, 11–22. doi:10.1017/S0033291799001452 [DOI] [PubMed] [Google Scholar]
  20. Kubzansky L. D., Davidson K. W., & Rozanski A (2005). The clinical impact of negative psychological states: Expanding the spectrum of risk for coronary artery disease. Psychosomatic Medicine, 67 (Suppl. 1), S10–S14. doi:10.1097/01.psy.0000164012.88829.41 [DOI] [PubMed] [Google Scholar]
  21. Lavie C. J., Milani R. V., & Ventura H. O (2009). Obesity and cardiovascular disease: Risk factor, paradox, and impact of weight loss. Journal of the American College of Cardiology, 53, 1925–1932. doi:10.1016/j.jacc.2008.12.068 [DOI] [PubMed] [Google Scholar]
  22. Lichtenstein P., De Faire U., Floderus B., Svartengren M., Svedberg P., & Pedersen N. L (2002). The Swedish Twin Registry: A unique resource for clinical, epidemiological and genetic studies. Journal of Internal Medicine, 252, 184–205. doi:10.1046/j.1365-2796.2002.01032.x [DOI] [PubMed] [Google Scholar]
  23. Lövdén M., Ghisletta P., & Lindenberger U (2005). Social participation attenuates decline in perceptual speed in old and very old age. Psychology and Aging, 20, 423–434. doi:10.1037/0882-7974.20.3.423 [DOI] [PubMed] [Google Scholar]
  24. Luppino F. S., de Wit L. M., Bouvy P. F., Stijnen T., Cuijpers P., Penninx B. W., & Zitman F. G (2010). Overweight, obesity, and depression: A systematic review and meta-analysis of longitudinal studies. Archives of General Psychiatry, 67, 220–229. doi:10.1001/archgenpsychiatry.2010.2 [DOI] [PubMed] [Google Scholar]
  25. McArdle J. J. (2001). A latent difference score approach to longitudinal dynamic structural analyses. In Cudeck R., duToit S., & Sorbom D. (Eds.), Structural equation modeling: Present and future (pp. 342–380). Lincolnwood, IL: Scientific Software International. [Google Scholar]
  26. McArdle J. J., Hamagami F., Meredith W., & Bradway K. P (2000). Modeling dynamic hypothesses of Gf-Gc theory using longitudinal life-span data. Learning and Individual Differences, 12, 53–79. doi:10.1016/S1041-6080(00)00036-4 [Google Scholar]
  27. McArdle J. J., Hamgami F., Jones K., Jolesz F., Kikinis R., Spiro A. III, & Albert M. S (2004). Structural modeling of dynamic changes in memory and brain structure using longitudinal data from the normative aging study. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, 294–304. doi:10.1093/geronb/59.6.P294 [DOI] [PubMed] [Google Scholar]
  28. McClearn G. E., Johansson B., Berg S., Pedersen N. L., Ahern F., Petrill S. A., & Plomin R (1997). Substantial genetic influence on cognitive abilities in twins 80 or more years old. Science (New York, N.Y.), 276, 1560–1563. doi:10.1126/science.276.5318.1560 [DOI] [PubMed] [Google Scholar]
  29. Muthén, L. K., & Muthén, B. O. (2012–2014). Mplus User’s Guide. Seventh Edition. Los Angeles, CA: Muthén & Muthén. [Google Scholar]
  30. Ohayon M. M. (2007). Epidemiology of depression and its treatment in the general population. Journal of Psychiatric Research, 41, 207–213. doi:10.1016/j.jpsychires.2006.10.006 [DOI] [PubMed] [Google Scholar]
  31. Onyike C. U., Crum R. M., Lee H. B., Lyketsos C. G., & Eaton W. W (2003). Is obesity associated with major depression? Results from the Third National Health and Nutrition Examination Survey. American Journal of Epidemiology, 158, 1139–1147. doi:10.1093/aje/kwg275 [DOI] [PubMed] [Google Scholar]
  32. Palinkas L. A., Wingard D. L., & Barrett-Connor E (1996). Depressive symptoms in overweight and obese older adults: A test of the “jolly fat” hypothesis. Journal of Psychosomatic Research, 40, 59–66. doi:10.1016/0022-3999(95)00542-0 [DOI] [PubMed] [Google Scholar]
  33. Pedersen N. L., & Reynolds C. A (1998). Stability and change in adult personality: Genetic and environmental components. European Journal of Personality, 12, 365–386. doi:10.1002/(SICI)1099-0984(1998090)12:5<365::AID-PER335>3.0.CO;2-N [Google Scholar]
  34. Peeters A., Barendregt J. J., Willekens F., Mackenbach J. P., Al Mamun A., & Bonneux L.; NEDCOM, the Netherlands Epidemiology and Demography Compression of Morbidity Research Group (2003). Obesity in adulthood and its consequences for life expectancy: A life-table analysis. Annals of Internal Medicine, 138, 24–32. doi:10.7326/0003-4819-138-1-200301070-00008 [DOI] [PubMed] [Google Scholar]
  35. Radloff L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. doi:10.1177/014662167700100306 [Google Scholar]
  36. Roberts R. E., Deleger S., Strawbridge W. J., & Kaplan G. A (2003). Prospective association between obesity and depression: Evidence from the Alameda County Study. International Journal of Obesity and Related Metabolic Disorders: Journal of the International Association for the Study of Obesity, 27, 514–521. doi:10.1038/sj.ijo.0802204 [DOI] [PubMed] [Google Scholar]
  37. Ross C. E. (1994). Overweight and depression. Journal of Health and Social Behavior, 35, 63–79. doi:10.2307/2137335 [PubMed] [Google Scholar]
  38. Sachs-Ericsson N., Burns A. B., Gordon K. H., Eckel L. A., Wonderlich S. A., Crosby R. D., & Blazer D. G (2007). Body mass index and depressive symptoms in older adults: The moderating roles of race, sex, and socioeconomic status. The American Journal of Geriatric Psychiatry: Official Journal of the American Association for Geriatric Psychiatry, 15, 815–825. doi:10.1097/JGP.0b013e3180a725d6 [DOI] [PubMed] [Google Scholar]
  39. Visser M., Pahor M., Taaffe D. R., Goodpaster B. H., Simonsick E. M., Newman A. B.,…Harris T. B (2002). Relationship of interleukin-6 and tumor necrosis factor-alpha with muscle mass and muscle strength in elderly men and women: The Health ABC Study. The Journals of Gerontology. Series A: Biological Sciences and Medical Sciences, 57, 326–332. doi:10.1093/gerona/57.5.M326 [DOI] [PubMed] [Google Scholar]
  40. Weiss A., Beloosesky Y., Boaz M., Yalov A., Kornowski R., & Grossman E (2008). Body mass index is inversely related to mortality in elderly subjects. Journal of General Internal Medicine, 23, 19–24. doi:10.1007/s11606-007-0429-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. de Wit L., Luppino F., van Straten A., Penninx B., Zitman F., & Cuijpers P (2010). Depression and obesity: A meta-analysis of community-based studies. Psychiatry Research, 178, 230–235. doi:10.1016/j.psychres.2009.04.015 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gbz022_suppl_Supplemental-Table-1
Click here for additional data file. (13KB, docx)
gbz022_suppl_Supplemental-Table-2
Click here for additional data file. (12.9KB, docx)

Articles from The Journals of Gerontology Series B: Psychological Sciences and Social Sciences are provided here courtesy of Oxford University Press

RESOURCES