SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

Objective

Emerging evidence suggests that psychosocial stress may influence weight gain. The relationship between stress and weight change and whether this was influenced by demographic and behavioral factors was explored.

Design and Methods

A total of 5,118 participants of AusDiab were prospectively followed from 2000 to 2005. The relationship between stress at baseline and BMI change was assessed using linear regression.

Results

Among those who maintained/gained weight, individuals with high levels of perceived stress at baseline experienced a 0.20 kg/m2 (95% CI: 0.07-0.33) greater mean change in BMI compared with those with low stress. Additionally, individuals who experienced 2 or ≥3 stressful life events had a 0.13 kg/m2 (0.00-0.26) and 0.26 kg/m2 (0.14-0.38) greater increase in BMI compared with people with none. These relationships differed by age, smoking, and baseline BMI. Further, those with multiple sources of stressors were at the greatest risk of weight gain.

Conclusion

Psychosocial stress, including both perceived stress and life events stress, was positively associated with weight gain but not weight loss. These associations varied by age, smoking, obesity, and multiple sources of stressors. Future treatment and interventions for overweight and obese people should consider the psychosocial factors that may influence weight gain.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

Obesity prevalence is increasing worldwide, and its contribution to morbidity and mortality from chronic disease is now well established [1]. Although the main drivers of obesity, such as over-nutrition and physical inactivity, are well characterized, evidence suggests that other factors also play a role in weight gain [2]. Psychosocial stress, for example, resulting from occupational, personal, or financial strain, has been suggested as a risk factor for weight gain [3]. Psychosocial stress may lead to weight gain through neuroendocrine and inflammatory pathways that directly increase abdominal adiposity [4]. Alternatively, stress could lead to the development of obesity through changes in health behaviors such as diet and physical activity [5]. Stress may affect food choices by reducing time available for food preparation and increasing preferences for high-fat energy-dense foods, therefore promoting positive energy balance [6]. Stress has also been shown to decrease participation in leisure-time physical activity [7].

Epidemiological evidence linking stress to weight gain has shown weak associations. A 2011 meta-analysis of 14 prospective studies revealed a weak positive relationship between stress (general life stress, caregiver stress and work stress) and objectively measured adiposity [8]. A key conclusion of this meta-analysis was the need to elucidate potential moderating variables of this relationship. Prior research has suggested that the effects of stress on weight gain may differ by sex [9, 10], baseline BMI [11, 12], and cortisol reactivity [13]. These factors may cause some people to gain more weight under stressful circumstances, whilst others may gain less weight or even lose weight when stressed [12]. However, the extent to which the association between stress and weight change differs according to demographic and other factors remains unclear.

Using a national, population-based sample of Australian adults, we aimed to explore the relationship between psychosocial stress and BMI change over 5 years. In addition, we aimed to investigate whether this relationship differed according to several demographic and behavioral characteristics.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

Study population

The baseline Australian Diabetes Obesity and Lifestyle study (AusDiab) was a national, population-based survey of adults (N = 11,247) aged ≥25 years in 1999–2000. Baseline methods and response rates have been described in detail elsewhere [14]. In brief, seven census collector districts were randomly selected from each of the six Australian states and the Northern Territory (total clusters = 42). Following a brief household interview, participants were invited to attend a biomedical examination that also included extensive interviewer-administered questionnaires. Among those who completed the household interview, 55.3% attended the biomedical examination. All eligible participants were invited to attend a follow-up assessment in 2004–2005. Among the eligible participants, 6,400 (60.0%) returned for the 5-year follow-up. Of these, we excluded participants with missing information on BMI, education, psychosocial stress, or health behaviors (smoking, alcohol, diet, and physical activity) (N = 1,282), leaving a study sample of 5,118. The study was approved by the Ethics Committees of the International Diabetes Institute and Monash University. Written informed consent was obtained from all participants.

Study variables

Demographic information

Information on age, sex, and education were collected by an interviewer-administered questionnaire as previously described [14, 15]. Education was classified into four categories based on responses to the question enquiring about the highest educational qualifications attained: 1) up to secondary school education; 2) trade/technical certificates; 3) associate degree, undergraduate diplomas, nursing/teaching qualifications; and 4) bachelor degree, post-graduate qualifications. At baseline and follow-up, the examination included blood samples and anthropometric measurements. Height, weight, and waist circumference were measured as described previously [16]. BMI was calculated as weight (kg) divided by height (m2) and categorized according to World Health Organisation (WHO) guidelines [17]. The outcome measure was continuous change in BMI between 2000 and 2005. Participants were categorized as “lost weight” if they lost more than 3% of baseline BMI, or “maintained/gained weight” if BMI change was within or greater than 3% of baseline BMI according to previously recommended cut-off points [18]. Participants were analyzed in these separate groups as it is known that stress can have differential effects on weight change, with some people gaining weight in response to stress, whilst others lose weight [11].

Health behaviors

The Active Australia questionnaire measured total leisure-time physical activity (including walking for transport) in minutes reported for the previous week [19]. Total physical activity time was calculated as the sum of time spent walking (if continuous and for >10 minutes) or performing moderate-intensity activity, plus double the time spent in vigorous-intensity activity. This double weighting has been used because of the need to reflect that participation in vigorous intensity physical activity confers greater health benefits than participation in moderate activity [20]. Participants were then categorized as meeting guidelines (≥150 min/week) or not meeting guidelines (≥0 min/week and <150 min/week) [14, 19].

Smoking history was assessed by questionnaire and dichotomized into smokers (current) and nonsmokers (never smoked and ex-smokers) [21]. Daily energy intake (kJ/day) and alcohol consumption (g/day) was assessed with the self-administered Anti-Cancer Council of Victoria food frequency questionnaire [22, 23]. Expected energy requirements (EERs), the estimated number of daily calories an individual requires in order to maintain his or her current weight, was determined using the Institute of Medicine equation (Appendix 1) based on an individual's, sex, age, height, weight and physical activity [24]. Using daily energy intake, participants were then categorized into above or below their EER.

Psychosocial stress

Perceived stress was measured at baseline using the Perceived Stress Questionnaire (PSQ)[25], comprising 30 items assessing perceptions of stress (e.g., You feel tense) over the past 12 months, with responses ranging from almost never to usually on a four-point Likert scale. The PSQ index was derived from the raw scores, ranging from 30 to 120 (higher scores reflecting elevated perceived stress). As these data were not normally distributed, they were categorized into quartiles. A more “objective” measure of stress was the life events scale in which stressful life events that had occurred in the preceding 12 months were also reported [17]. Thirteen items indicating different life stressors (such as marriage breakdown, financial hardship) were summed to provide a discrete score (0-13). Total scores were further categorized as follows: 0 = no stressful life events; 1 = one stressful life event; 2 = two stressful life events; 3 = three or more stressful life events.

Statistical analysis

Differences in baseline characteristics between participants were assessed using Pearson's chi-square test, t-tests, and one-way ANOVA as appropriate. The relationship between psychosocial stress and continuous BMI change in both categories of weight change (maintained/gained weight or lost weight) was assessed using linear regression. Two different models were fitted, with ptrends reported. Model 1 included age, sex, and education, while Model 2 additionally adjusted for health behaviors that may mediate the relationship between stress and weight change (smoking, alcohol, energy intake, and physical activity). Covariates included in Model 2 were included as continuous variables unless assumptions of normality were violated in which case the categorical variable of that measure was used instead. To understand the effect of moderating variables, we analyzed a multivariate model including all previously mentioned variables, baseline BMI, and both stress measures to determine which variables significantly predicted weight gain. Those that were significant were then stratified to observe trends within subgroups, adjusted as per Model 2. These subgroup analyses were restricted to those who had maintained or gained weight over the 5-year follow up (N = 4,413) as it is likely that people who lose weight have different behavioral patterns in response to stress compared with those who maintain or gain weight [26]. Additionally, we were underpowered to explore those who lost weight in subgroup stratifications. We also tested for interactions between psychosocial stress and all a priori potential moderating factors. Given the lack of statistical power inherent in interaction tests, we used a p-value cut point of p = 0.2 [27]. All analyses were conducted using Stata version 12 (StataCorp, College Station, TX, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

Baseline characteristics

Participants who returned for follow-up were more likely to have a higher educational attainment, higher physical activity levels, consume less alcohol, less likely to be a current smoker and had fewer stressful life events compared to those who did not return for follow-up (Appendix 2). Among the 5,118 participants (2,781 women and 2,377 men) included in this study, a mean BMI change of 0.81 (±1.93) kg/m2 was observed; range: −18.40 to 19.05 kg/m2. Higher perceived stress was observed in women and in those with younger age, obesity, higher energy intake, higher educational attainment, and less physical activity (Table 1). Participants with high perceived stress were also more likely to have experienced ≥3 stressful life events in the previous year. There were no differences in smoking status or daily alcohol intake across the quartiles of perceived stress. Similarly, those with higher numbers of stressful life events were more likely to be women, of younger age, obese, have a higher energy intake, smokers, more educated, undertake less physical activity, and were more likely to report high perceived stress than those with no stressful life events (Appendix 3).

Table 1. Participant characteristics at baseline by quartiles of perceived stress
CharacteristicQuartiles of perceived stressP-value
1234
  1. Data are means ± SD or proportions (95% CI).

  2. Physical activity and alcohol intake reported as median (25th, 75th percentiles).

Age (years)57.3 ± 12.750.7 ± 12.348.4 ± 11.346.7 ± 10.3<0.001
Sex (women)50.5 (47.9, 53.0)52.2 (49.5, 54.9)56.6 (53.7, 59.4)59.3 (56.5, 62.1)<0.001
Educational attainment    <0.001
Secondary school40.6 (38.1, 43.1)37.8 (35.1, 40.4)32.0 (29.3, 34.6)32.4 (29.7, 35.1)
Trade certificate31.4 (28.9, 33.8)30.3 (27.8, 32.8)31.5 (28.9, 34.2)27.0 (24.4, 29.5)
Associate, undergraduate diploma, etc.13.2 (11.4, 14.9)13.6 (11.7, 15.4)14.5 (12.4, 16.5)14.9 (12.9, 17.0)
Bachelor degree, post-graduate qualification14.8 (13.0, 16.7)18.4 (16.2, 20.4)22.1 (19.7, 24.4)25.7 (23.2, 28.2)
BMI group
Normal35.6 (33.2, 38.1)38.1 (35.4, 40.7)39.9 (37.1, 42.7)38.9 (36.1, 41.7)<0.01
Overweight45.6 (43.0, 48.2)40.1 (37.5, 42.8)39.0 (36.2, 41.8)39.3 (36.5, 42.1)
Obese18.8 (16.8, 20.8)21.8 (19.6, 24.1)21.1 (18.8, 23.5)22.0 (19.4, 24.2)
Physical activity (min/week)2210 (60, 480)180 (40, 420)150 (45, 380)120 (30, 328)<0.001
Current smoker10.9 (9.2, 12.5)11.8 (10.0, 13.5)10.9 (9.1, 12.7)12.7 (10.8, 14.6)0.42
Alcohol (g/day)27.7 (1.1, 20.5)8.7 (1.5, 21.3)8.1 (1.4, 21.2)6.8 (1.4, 19.2)0.34
Energy intake (kj/day)7,856 ± 2,9108,213 ± 3,1788,264 ± 3,2568,482 ± 3,417<0.001
≥3 stressful life events in previous 12 months6.3 (5.1, 7.6)16.8 (14.8, 18.9)29.2 (26.7, 31.8)51.6 (48.7, 54.4)<0.001

Relationship between psychosocial stress and BMI change

The relationship between psychosocial stress and BMI change is shown in Table 2. Among participants who maintained or gained weight over the follow-up period (mean BMI change 1.28 [±1.51] kg/m2; range −1.55 to 19.05 kg/m2), those who reported the highest quartile of perceived stress had a 0.20 kg/m2 (95% CI: 0.07-0.33) greater mean change in BMI compared with those in the lowest quartile of perceived stress, ptrend across quartiles = 0.004. The magnitude of this relationship did not change following adjustment for health behaviors. Stressful life events were also a significant predictor of BMI change, whereby 2 and ≥3 stressful life events were associated with a 0.13 kg/m2 (0.00-0.26) and 0.26 kg/m2 (0.14-0.38) greater mean increase in BMI during follow-up, respectively, compared with people who had not experienced a stressful life event in the past year, ptrend < 0.001. Adjustment for health behaviors did not appreciably change these estimates. No relationship was observed between stress and BMI change among people who lost weight over the follow-up period (mean BMI change −2.13 [±1.68] kg/m2; range −18.40 to −0.64 kg/m2). In a sensitivity analysis, additional adjustment in Models 1 and 2 for baseline BMI did not alter these results.

Table 2. Linear regression between psychosocial stress and BMI change
 Model 11Model 23
  1. Data are mean change in BMI (β-coefficient (95% CI).

  2. Model 1 adjusted for age, sex and education.

  3. Model 2 additionally adjusted for health behaviors (alcohol, smoking, energy intake, and physical activity).

  4. a

    p < 0.05.

People who lost weight (n = 705)
Quartiles of perceived stress
1 (low perceived stress)refref
20.04 (−0.29, 0.36)−0.02 (−0.36, 0.33)
30.24 (−0.12, 0.61)0.30 (−0.08, 0.69)
4 (high perceived stress)0.23 (−0.15, 0.60)0.26 (−0.13, 0.65)
No. of stressful life events in previous 12 months
0refref
10.09 (−0.24, 0.42)0.14 (−0.20, 0.49)
2−0.10 (−0.46, 0.25)−0.08 (−0.45, 0.29)
≥30.16 (−0.20, 0.51)0.27 (−0.09, 0.64)
People who maintained/gained weight (n = 4413)
Quartiles of perceived stress
1 (low perceived stress)refref
2−0.03 (−0.16, 0.09)−0.03 (−0.16, 0.09)
3−0.03 (−0.15, 0.10)−0.02 (−0.15, 0.10)
4 (high perceived stress)0.20 (0.07, 0.33)a0.20 (0.07, 0.33)a
No. of stressful life events in previous 12 months
0refref
10.04 (−0.08, 0.15)0.07 (−0.05, 0.19)
20.13 (0.00, 0.26)a0.13 (−0.00, 0.27)
≥30.26(0.14, 0.38)a0.27 (0.15, 0.40)a

Moderators of the psychosocial stress and BMI change relationship in people who maintained/gained weight

In multivariate regression analysis examining potential predictors of weight gain, those that independently predicted weight gain were smoking, energy intake, sex, age, baseline BMI, perceived stress, and stressful life events (Appendix 4). Models were then stratified by these variables to observe trends within subgroups.

Smoking

Among nonsmokers, those with the highest quartile of perceived stress had a 0.25 kg/m2 (0.11-0.39) greater mean increase in BMI during follow up compared with those of the lowest quartile of stress, ptrend < 0.001 (Table 3). Similar patterns were seen in nonsmokers who had experienced 2 or ≥3 stressful life events. These associations were not observed in smokers, with a significant interaction evident between smoking status and each of perceived stress and stressful life events, p < 0.01 and p < 0.05, respectively.

Table 3. Linear regression between psychosocial stress and BMI change in maintainers/gainers, stratified by various factors
SmokingCurrent smoker (n = 517)Nonsmoker (n = 3,896)P-value for interaction
  1. Data are mean change in BMI (β-coefficient (95% CI).

  2. a

    EER = expected energy requirement.

  3. Age was dichotomized based on mean baseline age.

  4. b

    p < 0.05.

Quartiles of perceived stress
1 (low perceived stress)refref<0.01
2−0.23 (−0.65, 0.18)−0.02 (−0.15, 0.11)
3−0.55 (−0.99, −0.11)b0.05 (−0.09, 0.18)
4 (high perceived stress)−0.01 (−0.41, 0.43)0.25 (0.11, 0.39)b
No. of stressful life events in previous 12 months
0refref<0.05
1−0.18 (−0.58, 0.23)0.10 (−0.03, 0.22)
2−0.20 (−0.564, 0.24)0.17 (0.03, 0.31)b
≥3−0.08 (−0.48, 0.32)0.232 (0.19, 0.45)b 
Energy intakeBelow EERa (n = 3,846)Above EER (n = 1,272) 
Quartiles of perceived stress
1 (low perceived stress)refref0.37
2−0.06 (−0.20, 0.08)0..08 (−0.18, 0.34)
3−0.04 (−0.19, 0.11)0.11 (−0.154, 0.37)
4 (high perceived stress)0.26 (0.11, 0.41)b0.16 (−0.10 – 0.428)
No. of stressful life events in previous 12 months
0refref0.35
10.07 (−0.07, 0.21)0.08 (−0.16, 0.323)
20.14 (−0.02, 0.29)0.186 (−0..08, 0.44)
≥30.23 (0.09, 0.38)b0.43 (0.19, 0.67)b 
SexMen (n = 2,041)Women (n = 2,372) 
Quartiles of perceived stress
1 (low perceived stress)refref0.23
20.00 (−0.15, 0.16)0.07 (−0.26, 0.13)
3−0.01 (−0.17, 0.15)−0.02 (−0.21, 0.18)
4 (high perceived stress)0.16 (−0.01, 0.33)0.27 (0.07, 0.47)b
No. of stressful life events in previous 12 months
0refref0.41
10.01 (−0.14, 0.15)0.11 (−0.07, 0.30)
20.12 (−0.05, 0.29)0.14 (−0.07, 0.34)
≥30.28 (0.12, 0.44)b0.26 (0.04, 0.46)b 
Ageb<50 years (n = 2,283)>50 years (n = 2,130) 
Quartiles of perceived stress
1 (low perceived stress)refref0.33
2−0.05 (−0.26, 0.17)−0.01 (−0.15, 0.14)
3−0.05 (−0.26, 0.16)0.05 (−0.12, 0.218)
4 (high perceived stress)0.25 (0.04, 0.45)b0.17 (−0.00, 0.35)
No. of stressful life events in previous 12 months
0refref0.13
10.13 (−0.06, 0.32)−0.02 (−0.13, 0.16)
20.16 (−0.04, 0.37)0.11 (−0.06, 0.27)
≥30.34 (0.15, 0.53)b0.20 (−0.01, 0.35) 
Perceived stressLow perceived stress (n = 2,342)High perceived stress (n = 2,071) 
No. of stressful life events in previous 12 months
0refref0.15
10.10 (−0.03, 0.23)0.02 (−0.21, 0.26)
20.01 (−0.15, 17)0.23 (0.01, 0.46)b
≥30.12 (−0.07, 0.31)0.34 (0.13, 0.55)b

Energy intake

Among participants with an energy intake below their EER, those with the highest quartile of perceived stress or those who experienced ≥3 stressful life events experienced a greater mean change in BMI relative to those with low perceived stress (0.26 kg/m2; 0.11-0.41), ptrend < 0.001, or no life events (0.23 kg/m2; 0.09-0.38), ptrend < 0.01. Among participants with an energy intake exceeding their EER, those with ≥3 stressful life events experienced a greater mean change in BMI relative to those with no life events (0.43 kg/m2, 0.19-0.67), but no significant association was seen between high perceived stress and weight gain for this subgroup. No significant interactions were observed between either stress measure and energy intake above or below EER.

Sex

Among women, those with the highest quartile of perceived stress experienced a 0.27 kg/m2 (0.07-0.47) greater mean change in BMI relative to those with low perceived stress, ptrend < 0.01.This relationship was not significant in men but there was no evidence of an interaction between perceived stress and sex. Both women and men who had experienced ≥3 stressful events had a 0.26 kg/m2 (0.07-0.46), ptrend < 0.01, and 0.28 kg/m2 (0.12-0.44), ptrend < 0.001, greater mean change in BMI, respectively, compared with those who had no stressful life events. There were no significant differences observed between men and women.

Age

Younger adults (<50 years) who experienced ≥3 stressful life events had a greater mean change in BMI compared with those who had no life events (0.34 kg/m2 [0.15-0.53]), and this effect was greater than in older adults (>50 years) who experienced a mean change in BMI of 0.21 kg/m2 (0.04-0.38). The relationship between perceived stress and BMI change did not differ significantly by age group.

Baseline BMI

Among those with normal BMI at baseline, those who had the highest quartile of perceived stress experienced a 0.24 kg/m2 (0.06-0.43) greater mean change in BMI compared with those who had low perceived stress, ptrend < 0.01 (Table 4). This relationship was not significant in those who were overweight or obese at baseline, but no interaction was observed between baseline BMI and perceived stress. Individuals who had a normal BMI and ≥3 stressful life events at baseline had a 0.21 kg/m2 (0.04-0.38) greater mean change in BMI compared with those who had none, ptrend < 0.01. In those who were overweight at baseline, 2 and ≥3 stressful life events was associated with a greater mean change in BMI compared with no stressful life events: (0.23 kg/m2, 95% CI: 0.02-0.45, and 0.28 kg/m2, 95% CI: 0.08-0.46), ptrend < 0.01. These relationships were not significant in those who were obese at baseline. A significant interaction was evident between baseline BMI group and stressful life events, p = 0.13.

Table 4. Linear regression between psychosocial stress and BMI change in maintainers/gainers, stratified by baseline BMI group
 Normal (n = 1,738)Overweight (n = 1,815)Obese (n = 860)P value for interaction
  1. Data are mean change in BMI (β-coefficient (95% CI).

  2. a

    p < 0.05.

Quartiles of perceived stress
1 (low perceived stress)refrefref0.37
20.01 (−0.17, 0.18)−0.04 (−0.23, 0.16)−0.20 (−0.54, 0.15)
30.06 (−0.12, 0.24)−0.05 (−0.25, 0.16)−0.22 (−0.58, 0.14)
4 (high perceived stress)0.24 (0.06, 0.43)a0.18 (−0.03, 0.39)−0.03 (−0.30, 0.43)
No. of stressful life events in previous 12 months
0refrefref0.13
10.04 (−0.13, 0.120)0.07 (−0.11, 0.26)−0.05 (−0.29, 0.39)
20.11 (−0.07, 0.30)0.23 (0.02, 0.45)a−0.25 (−0.61, 0.11)
≥30.21 (0.04, 0.38)a0.28 (0.08, 0.46)a0.16 (−0.15, 0.51)

High and low perceived stress

To investigate whether the association between the number of stressful life events experienced and weight gain differed according to perceived stress, the analysis was stratified by high and low levels of perceived stress (Table 3). Among those who had high perceived stress, a greater number of stressful life events was associated with a greater change in BMI. Among those who had low perceived stress, there was no relationship between the number of stressful life events and change in BMI. There was a significant interaction between level of perceived stress and stressful life events, p = 0.15.

We also investigated the inclusion of both stress markers in the same model to elucidate which measure was more important in the prediction of weight gain in this group of people (Appendix 4). In fully adjusted models, ≥3 stressful life events were associated with a greater increase in weight gain (0.22 kg/m2; [0.08, 0.36]) compared to those with no life events, independent of one's perceived stress. However, in the same model, high perceived stress was not associated with an increased risk of weight gain relative to low perceived stress when stressful life events was also taken into account (0.01 kg/m2; [−0.02, 0.27]).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

This study has shown that psychosocial stress, both perceived stress and stressful life events, is positively associated with weight gain, but not weight loss, over a 5-year period. Additionally, we demonstrated that associations with weight gain are, to some extent, influenced by demographic and behavioral factors: the effects were stronger if participants were normal weight or overweight, nonsmokers, or younger. Furthermore, stressful life events was a stronger predictor of weight gain compared to perceived stress, independent of covariates; however, the risk of weight gain was greatest in those with both high perceived stress and a number of stressful life events.

These findings extend earlier cross-sectional and longitudinal research examining the association between psychosocial stress and weight gain by attempting to elucidate potential moderating factors involved. Wardle et al. [8] analyzed 14 longitudinal cohort studies from the United States, United Kingdom, Europe, and Japan to show that psychosocial stress was a risk factor for weight gain, although effects were only small and cannot be directly compared to our results because of the variability in analyses between studies. Additionally, contrary to our results, they concluded that effects were stronger in men. Inconsistent data exist concerning sex differences in the relationship between stress and weight gain, with some studies reporting a greater effect in women [28-30] and others in men [11, 31]. These inconsistencies may be because of the different measures of stress across studies as previous research has suggested that different types of stressors may differentially influence weight gain in men and women [31-33]. Stressors external to work or finances, such as neighborhood stress and/or strain in relationships with family have been associated with weight gain in women, but not men [28]. Work stress, however, may be associated with weight gain in men but not women [29]. We did not find significant differences between men and women. This may be explained by the stress measures used, because they related to general life stressors that could be said to be similar for both men and women. It is also noteworthy that we did not see a relationship between stress and weight loss, though this has been shown in prior studies [11, 34]. It is possible that our results differ to previous literature as we were underpowered to detect small effect sizes.

We have shown for the first time that a stressful life event appears to be a significant predictor of weight gain, independent of perceived stress. However, those with both high perceived stress and a number of stressful life events were at the highest risk of weight gain with the level of perceived stress significantly moderating the relationship between stressful life events and weight gain. Though this has not been shown before, it is perhaps not surprising that multiple sources of high stress synergistically increase your risk of weight gain. A similar phenomenon has been found with regard to the development of depression, with a study showing that perceived stress moderates the association between negative life events and depression such that in those with low perceived stress, negative life changes had only a minimal impact [2]. It is therefore important that when predicting the risk of weight gain, multiple markers of stress should be considered.

We found that nonsmokers, but not smokers, had a significant risk of weight gain if highly stressed. Other studies of stress and smoking report that stress is associated with greater tobacco use [29, 35] and that tobacco use is associated with weight loss, rather than gain [36]. Leventhal et al. [37] explored the role of tobacco use as a moderator of the association between depression and obesity, and found that nonsmokers experienced a greater increase in weight as a result of depression, compared with smokers. It is therefore possible that smokers in this cohort of Australian adults increased their smoking in response to stress, offsetting the effects of weight gain, whereas the nonsmokers may have sought other behavioral strategies to deal with stress, such as increased sedentary time and/or increased eating. This emerging body of evidence suggests that smoking status should be considered when understanding how behavioral mechanisms interact within relationships between psychological risk factors and weight gain, and also which individuals may benefit most from obesity interventions that also target psychosocial stress.

Previous research has also examined baseline BMI as an effect modifier of the relationship between stress and weight gain [11, 12]. The Midlife in the United States study found that psychosocial stress was associated with greater weight gain in men and women with a higher baseline BMI over 9 years [12], whereas a prospective cohort of British civil servants observed this relationship in men but not women [11]. Our results differ somewhat in that we observed these associations in those who were normal or overweight at baseline, but not obese, and only for stressful life events, not perceived stress. Furthermore, we found that stressful life events, but not perceived stress, appeared to have a greater impact on the risk of weight gain in younger people, which is consistent with previous literature [12, 16].

We did not find that diet, measured by daily energy intake, or physical activity could explain the weight gain in those people with high stress levels. Although there are no similar studies exploring the moderating effects of diet and physical activity on the relationship between stress and weight gain, many studies indicate that the relationship between stress and weight gain does not appreciably change when adjusting for these health behaviors [32, 38]. Our results support this, suggesting that energy intake and physical activity may not be in the causal pathway between stress and weight gain. However, because of the self-reporting nature of these variables, it is more likely that we, and others, may have not been able to detect a real result because of measurement error.

The key strength of this study is its prospective design and ability to establish a temporal relationship between psychosocial stress and weight gain. Data were available on two markers of stress allowing discrimination of the independent effects of perceived stress and stressful life events on weight gain.

This study has several potential limitations. First, although AusDiab is a national, population-based study, the 55.3% response rate suggests that this sample may not be wholly representative of all Australians [14]. Furthermore, we have shown that those who returned for follow-up were, on average, healthier than those who did not return for follow-up. We expect that this would have led to an underestimation of our observed associations between psychosocial stress and weight gain. Second, self-reported health behaviors, particularly diet and physical activity, are notoriously difficult to measure accurately [39]. This may have led to an underestimation or a biased effect that these health behaviors play in influencing the relationship between stress and weight gain. Additionally, though not a focus of this study, it is possible that other factors such as inflammatory markers may play a role in the relationship between stress and weight gain. However, data on key inflammatory markers were not collected and we were therefore not able to explore these pathways further.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

In this population-based sample of Australian adults, psychosocial stress, both as perceived and life events, was associated with weight gain but not weight loss. These associations varied by age, smoking, and weight status. Further, the occurrence of a stressful life event increased the risk of weight gain, independent of perceived stress, though the greatest impact on weight gain was evident in those with both high perceived stress and numerous stressful life events. Future research is warranted to further elucidate the extent to which these factors contribute to the complex relationship between stress and weight gain. The interaction between different sources of psychosocial stress should also be explored in further detail. Future treatment and interventions for overweight and obese people should consider the psychosocial factors that may influence weight gain above and beyond the traditional biological pathways.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

The AusDiab study co-coordinated by the Baker IDI Heart and Diabetes Institute gratefully acknowledges the generous support given by National Health and Medical Research Council (NHMRC grant 233200), Australian Government Department of Health and Ageing. Abbott Australasia Pty Ltd, Alphapharm Pty Ltd, AstraZeneca, Bristol-Myers Squibb, City Health Centre-Diabetes Service-Canberra, Department of Health and Community Services – Northern Territory, Department of Health and Human Services – Tasmania, Department of Health – New South Wales, Department of Health – Western Australia, Department of Health – South Australia, Department of Human Services – Victoria, Diabetes Australia, Diabetes Australia Northern Territory, Eli Lilly Australia, Estate of the Late Edward Wilson, GlaxoSmithKline, Jack Brockhoff Foundation, Janssen-Cilag, Kidney Health Australia, Marian & FH Flack Trust, Menzies Research Institute, Merck Sharp & Dohme, Novartis Pharmaceuticals, Novo Nordisk Pharmaceuticals, Pfizer Pty Ltd, Pratt Foundation, Queensland Health, Roche Diagnostics Australia, Royal Prince Alfred Hospital, Sydney, Sanofi Aventis, sanofi-synthelabo, and the Victorian Government's OIS Program. Also, for their invaluable contribution to the set-up and field activities of AusDiab, we are enormously grateful to A Allman, B Atkins, S Bennett, A Bonney, S Chadban, M de Courten, M Dalton, D Dunstan, T Dwyer, H Jahangir, D Jolley, D McCarty, A Meehan, N Meinig, S Murray, K O'Dea, K Polkinghorne, P Phillips, C Reid, A Stewart, R Tapp, H Taylor, T Whalen, and F Wilson.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

APPENDIX 1: Calculation of Expected Energy Requirements (EER)

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

Adult Men: EER = (662 – (9.53 × Age)) + PA ((15.91 × wt) + (539.6 × ht))

Adult women: EER = (354 – (6.91 × Age)) + PA (9.36 × wt) + (726 × ht)),

where wt = weight, PA = physical activity, and ht = height. Physical activity coefficients were determined using guidelines outlined in the table below.

Table 5. 
Physical activity coefficients for the calculation of EER
 Sedentary (0 min)Moderately active (>0 and <150 min)Active (≥420 and <1260 minutes)Very active (>1,260 min)
Adult men11.111.251.48
Adult women11.121.271.45

APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years
Table 6. 
Baseline characteristicReturned (n = 5118)Not returned (n = 3669)P-value
  1. Data are means ± SD or proportions; only participants with complete demographic and health behavior information at baseline were included in this analysis.

Age (years)51.07 ± 12.4450.61 ± 15.930.126
Sex (% women)54.3654.540.867
Educational attainment (%)
Secondary school35.9944.54<0.001
Trade certificate30.1331.07
Associate, undergraduate diploma, etc.13.9710.17
Bachelor degree, post-graduate qualification19.9114.23
BMI26.84 ± 4.7627.04 ± 5.140.057
Physical activity (min/week)283 ± 332267 ± 333<0.01
Current smoker (%)11.5321.86<0.001
Alcohol (g/day)14.21 ± 17.77123.95 ± 18.15<0.01
Energy intake (kj/day)8,187 ± 3,1898,216 ± 3,6990.699
≥3 stressful life events in previous 12 months (%)24.7627.310.05
High perceived stress (%)22.9423.900.433

APPENDIX 3: Participant characteristics at baseline by number of stressful life events

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years
Table 7. 
CharacteristicNumber of stressful life eventsP-value
No stressful life events1 stressful life event2 stressful life events≥3 stressful life events
  1. Data are means ± SD or proportions (95% CI); *physical activity and alcohol intake reported as median (25th, 75th percentiles).

Age (years)52.6 ± 13.452.8 ± 12.650.4 ± 12.047.9 ± 10.6<0.001
Gender (% women)686 (48.7)779 (53.5)547 (55.4)769 (60.7)<0.001
Educational attainment
Secondary school542 (38.5)572 (39.3)337 (34.1)391 (30.9)<0.001
Trade certificate437 (31.0)435 (29.9)282 (28.5)388 (30.6)
Associate, undergraduate diploma, etc.161 (11.4)181 (12.4)164 (16.6)209 (16.5)
Bachelor degree, post-graduate qualification268 (19.0)267 (18.4)205 (20.8)279 (22.0)
BMI group
Normal575 (29.6)541 (18.1)351 (18.1)478 (24.6)<0.01
Overweight582 (27.6)625 (29.7)403 (19.1)498 (23.6)
Obese251 (23.6)289 (27.1)234 (22.0)291 (27.3)
Physical activity (min/day)*180 (50, 420)180 (40, 420)180 (30, 420)150 (40–360)<0.001
Current smoker (%)149 (10.6)153 (10.5)110 (11.1)178 (14.1)<0.05
Alcohol (g/day)*7.7 (1.3, 20.5)8.6 (1.3, 20.3)8.4 (1.5, 21.1)6.5(1.1, 20.1)0.54
Energy intake (kJ/day)84.7 ± 13.685.3 ± 14.384.6 ± 14.384.1 ± 13.8<0.05
High perceived stress108 (7.7)201 (13.8)262 (26.5)608 (45.0)<0.001

APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgments
  9. References
  10. APPENDIX 1: Calculation of Expected Energy Requirements (EER)
  11. APPENDIX 2: Differences in baseline characteristics between returning and nonreturning participants
  12. APPENDIX 3: Participant characteristics at baseline by number of stressful life events
  13. APPENDIX 4: Predictors of weight gain in people who maintained or gained weight over 5 years
Table 8. 
 Coefficient (95% CI)
  1. a

    p < 0.05.

Perceived stress
1 (low perceived stress)ref
2−0.08 (−0.20, 0.04)
3−0.08 (−0.21, 0.05)
4 (high perceived stress)0.12 (−0.03, 0.26)
No. stressful Life events
0ref
10.06 (−0.07, 0.16)
20.09 (−0.05, 0.22)
≥30.19 (0.06, 0.33)a
Sex
Menref
Women0.36 (0.26, 0.47)a
Age (cont)−0.02 (−0.02, −0.02)a
Education
Secondary schoolref
Trade certificate−0.07 (−0.19, 0.04)
Associate, undergraduate diploma, etc.−0.16 (−0.30, −0.02)a
Bachelor degree, post-graduate qualification−0.15 (−0.28, −0.02)a
Smoking status
Currentref
Ex/non0.29 (0.15, 0.43)a
Energy intake (cont)−0.00 (−0.00, −0.00)a
Physical activity (sufficient)−0.02 (−0.11, 0.08)
Alcohol (cont)0.02 (−0.05, 0.08)
Baseline BMI
Normalref
Overweight0.18 (0.08, 0.28)a
Obese0.48 (0.35, 0.60)a