SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

Inconsistent results exist for whether or not weight cycling (WgtC) and weight variability (WgtV) increase mortality risk. The aim of this study was to examine the effect of WgtC and WgtV during adulthood on mortality risk. Data was obtained from the Women's Health Initiative (WHI) observational study (OS) dataset, acquired from the National Heart, Lung and Blood Institute (N = 47,473 overweight and obese women; age 50–79 years). Women were categorized (stable; WgtV: weight-gainer or loser; or WgtC) based on weight changes during early (18–35 years), mid (35–50 years), and late (50 years to current age) adulthood. Those with weight changes of <5% during all three time-periods were classified as being stable-weight. Weight-gainers were those with at least one period of weight-gain (≥5%) without a period of weight-loss (≥5%), and weight-losers were those with at least one period of loss without a period of gain during all time-periods. Those who experienced both a period of weight-gain and loss (≥5%) were categorized as WgtC. Compared to stable-weight individuals, WgtC and WgtV across adulthood were not significantly associated with mortality risk when the age-period of weight change was not considered. However, when considering the age period, increased mortality risk was observed for every 5 kg of weight-gain during early (hazard ratio (HR) = 1.04 (1.00–1.07)) or mid-adulthood (HR = 1.05 (1.02–1.08)), or for every 5 kg of weight-loss since mid (HR = 1.12 (1.01–1.24)) or late-adulthood (HR = 1.12 (1.04–1.20)). In conclusion, merely investigating WgtC and WgtV by weight changes across adulthood may not be sufficient to fully describe mortality risk, and the age at which the weight change occurred might be as important to consider.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

It is usually recommended for obese individuals to lose weight, as weight loss has been demonstrated to improve conditions associated with obesity such as hypertension, hyperlipidemia, type 2 diabetes, and cardiorespiratory diseases (1). Although weight loss is often attempted, it is rarely maintained once achieved, and regain is common (2). This brings forth concerns regarding the issue of weight cycling (WgtC) and weight variability (WgtV). In its simplest form, WgtC is defined as weight changes in opposite directions, that is, a weight loss followed by a weight gain, or vice versa. The more general term WgtV can be described as an interval of net weight loss or an interval of net weight gain across a defined time period in an individual's life.

The relationships between WgtC, WgtV and mortality risk have been inconsistent across studies, with deleterious effects reported in some (3,4,5), but not others (6,7,8). Studies indicating an association between WgtC, WgtV and increased mortality risk have had short follow-up periods (4), small sample sizes (3), or have attributed their findings to pre-existing diseases or unhealthy participant characteristics (9). Furthermore, the majority of these studies have only examined men (3,4,6). As ∼60% of adult women in the United States are overweight or obese (10) and an even larger proportion of women attempt weight loss (11) (resulting in more weight changes as compared to men (12)), it is important to further examine the consequences of weight changes in overweight and obese female populations.

Furthermore, the associations between weight gain or weight loss and mortality risk may differ according to the timing of weight change. For instance, weight gain during early to mid-adulthood is associated with increased disease and mortality risk (13), whereas the relationship between weight loss and mortality risk is not as clear. Among older individuals, weight loss is often associated with increased mortality risk (14), whereas weight maintenance is associated with lower mortality risk than weight loss or weight gain amongst overweight or obese (15). Previous studies do not take into account the ages at which weight change occurs, limiting their ability to fully describe relationships between WgtC, WgtV, and mortality risk. To our knowledge, no study has examined both the associations of WgtC and WgtV across adulthood as compared to associations of weight change during specific age periods with mortality risk. Thus, the purpose of the present investigation is to examine the influence of WgtC and WgtV throughout adult life and between different time-periods on mortality risk in a large sample of overweight and obese women.

Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

The Women's Health Initiative

Complete details regarding the Women's Health Initiative (WHI) have been published (16,17,18,19). Briefly, the WHI commenced in 1992 with the purpose of examining common causes of morbidity and mortality, such as cancer, osteoporosis, and cardiovascular disease (CVD) in postmenopausal women (17). In total, the original WHI study examined 161,808 women aged 50–79 years who were enrolled between 1993 and 1998. Recruitment was conducted based out of 40 clinical centres throughout the United States (18). The WHI consisted of two components: a clinical trial component, and an observational study (OS) component. Women who were unwilling to enroll or were ineligible for the clinical trial components were invited to enroll in the OS (n = 93,676). Participants were then followed for ∼8 years for changes in various characteristics and health habits. Partial access to the OS portion of the WHI database was obtained from the National Heart, Lung and Blood Institute (NHLBI) and was used for this analysis. All participants gave informed consent prior to participation in the study. Ethics approval was obtained for secondary data analysis for the present study through the York University Research Ethics Board.

Study sample

Exclusion criteria included: underweight or normal weight individuals (BMI <25.0 kg/m2; n = 28,974), age <50 or >79 years (n = 4), and those with missing data for the exposure (weights at 18 years, 35 years, 50 years, current), or baseline disease status (CVD, cancer, diabetes) (n = 17,225). Extreme values were individually assessed and excluded if appropriate. This left a final sample size of 47,473 overweight and obese women with complete data for the current analysis.

Measurement of exposure, covariate and outcome variables

During the baseline visit, trained personnel obtained height and weight measurements to the nearest 0.1 cm and 0.1 kg respectively. BMI was then calculated as weight in kilograms (kg) per meter (m) squared. Participants' weights at ages 18, 35, and 50 years (while not pregnant) were obtained by self-report, and were used to calculate the exposure variables of weight change (described later). Covariates included age, ethnicity (Asian or Pacific Islander; black or African American; white; other), income (low: <$20,000; middle: $20,000 to <$75,000; high: ≥$75,000), education (high school diploma or lower; college graduate/baccalaureate degree; some post graduate/professional degree), alcohol intake (none, past, moderate drinker: 1 to <7 drinks/week; heavy drinker: ≥7 drinks/week), hormone replacement therapy (never, past, current), smoking status (never, past, current), physical activity level (inactive: ≤7.5 metabolic equivalent (MET) h/week; moderately active: >7.5 to ≤21 MET h/week; active: >21 MET h/week) and baseline morbidity status (CVD, diabetes, and cancer) and were obtained by interview or by self-report using a standardized questionnaire. Outcome variable information for the final sample was obtained from mortality follow-up data over 7.0 ± 1.6 years, including adjudication of the cause of death.

Weight cycling and weight variability across adult life

To examine WgtC and WgtV across adulthood, participants were divided into categories (stable weight; WgtV: weight gainer or weight loser; and WgtC) based on their weight change between the defined age points (ages 18–35 years, 35–50 years and 50 years to current age). WgtV was calculated as the difference in body weight between the age points. Research has demonstrated that modest weight loss of 5–10% is sufficient to influence health (1,20). Thus, those who had a change of <5% of their body weight across all age points were classified as having stable weight. Weight gainers were those who gained ≥5% body weight without losing ≥5% body weight between any of the age points, and weight losers were those who lost weight without gaining weight between any of the age points. Those who experienced both a period of weight gain and a period of weight loss were categorized as WgtC. The patterns of weight change during the different age points as well as weight gain and loss experienced by those with WgtC are presented in Table 1.

Table 1.  Patterns of weight change throughout adulthood and between three time periods (18–35 years, 35–50 years, 50 years to current age)
inline image

Furthermore, WgtV during specific age periods was also assessed in order to isolate potential associations of weight change at different ages with future mortality risk. Participants were divided into categories of weight gain or weight loss between 18–35 years of age, 35–50 years of age, 18 to current age, 35 to current age, and 50 to current age, and the associations between each age period with mortality risk were individually assessed.

Statistical analyses

Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for the associations between WgtC, WgtV, and mortality risk during 7.0 ± 1.6 years of follow-up. WgtC and WgtV during adulthood were examined using the previously described categories with stable weight as the reference group in one Cox model. Separate Cox models were then generated to examine either weight gain or weight loss as continuous variables between 18–35 years of age, 35–50 years of age, 18 to current age, 35 to current age, and 50 to current age. Individuals with zero weight change were not included in these analyses.

The influence of WgtC and WgtV on mortality risk was examined using three models: crude models; Model 1-adjusted for BMI, age, ethnicity, income, education, alcohol intake, smoking status, physical activity level, and hormone replacement therapy; Model 2-further adjustment for pre-existing diseases at baseline (CVD, diabetes, and cancer). All statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC). Statistical significance was set at α = 0.05.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

Participant characteristics at baseline, including prevalent diseases and follow-up data are presented in Table 2. Briefly, women in this sample were mainly white (82%), of middle income (65%), and had a post secondary degree (49%). At baseline, CVD was present in 20.3% (n = 9,637), cancer in 13.4% (n = 6,364), and diabetes in 7.8% (n = 3,683) of participants. During an average of 7.0 ± 1.6 years of follow-up there were 3,192 deaths.

Table 2.  Baseline characteristics of study participants by weight change status, the Women's Health Initiative study (n = 47,473)
inline image

Influence of weight changes on mortality risk

In comparison to being weight stable, weight gain, weight loss, and WgtC across adult life were not significantly associated with mortality risk when the age-period during which the weight change occurred was not considered (Table 3). The HRs for weight changes during specific age periods (expressed per 5 kg of weight gain or loss) are shown in Figure 1. Adjustment for pre-existing diseases (model 2) attenuated or abolished some of the associations between WgtV and mortality risk.

Table 3.  Mortality risk according to weight change categories for weight changes across adulthood (n = 47,473)
inline image
image

Figure 1. Mortality risks for weight gain or loss during specific age periods (n = 47,473). Model 1: adjusted for BMI, age, ethnicity, income, education, alcohol intake, hormone replacement therapy, smoking status, and physical activity. Model 2: model 1 + adjustment for pre-existing diseases (cardiovascular disease, diabetes, and cancer). HRs are presented as per 5 kg weight gain or loss during each age interval. CI, confidence interval; HR, hazard ratio. *P < 0.05.

Download figure to PowerPoint

Weight gain and mortality risk

After adjustment for pre-existing diseases, the associations of weight gain between 18–35 years of age (HR = 1.04 (1.00–1.07)) and 35–50 years of age (HR = 1.05 (1.02–1.08)) remained significantly associated with increased mortality risk. A slight positive, albeit nonsignificant (HR = 1.02 (1.00–1.04), P = 0.052) association was present for weight gain between 18 to current age after adjustments for pre-existing diseases, whereas weight gain between 35 to current age, and 50 to current age was no longer significantly associated with mortality risk. Hence, these results indicate that weight gain during early and mid-adulthood may be more detrimental to longevity than weight gain later in life.

Weight loss and mortality risk

The association between weight loss and mortality risk was dependent on the age period at which the weight loss occurred. Weight loss during 18–35 years of age was not significantly associated with mortality (P > 0.05), whereas a slight, although nonsignificant (HR = 1.09 (1.00–1.19), P = 0.055) increase in mortality risk was observed in those reporting weight loss during 35–50 years of age, an effect that persisted even after adjustment for pre-existing diseases. Weight loss during 18 to current age was not significantly associated with mortality risk with and without adjustment for covariates, whereas women who experienced weight loss between 35 years to current age (HR = 1.12 (1.01–1.24)) and 50 years to current age (HR = 1.12 (1.04–1.20)) had significantly increased mortality risk, and adjustment for pre-existing diseases attenuated the magnitude of these associations. Thus, these results suggest that weight loss later in life may have a greater detrimental effect on mortality risk.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

WgtC and WgtV during adulthood as assessed by the overall weight change across adult life, without regard for the age at which the weight change occurred, were not significantly associated with mortality risk. Conversely, we observed that WgtV between certain age periods during adulthood may have differing associations with mortality risk. Our study has demonstrated that not only is the magnitude of weight gain and loss important to consider when investigating WgtC and WgtV, but the age at which an individual gains or loses weight should also be taken into consideration.

Numerous studies (6,9,21,22) utilize weight changes between multiple time points to investigate the relationships between WgtC, WgtV, and health risk. However, they fail to account for the age at which the weight was gained or lost. We demonstrate that the age at which the weight changes occur may have differential influences with mortality risk. Without consideration for the age at which the weight changes occurred in fact masked significant associations between weight changes and mortality risk. Other studies have demonstrated that weight gain during early and mid-adulthood is associated with increased disease and mortality risk (13,23). In the current study, it was observed that only weight gain during early or mid-adulthood, and not in late adulthood, was related to increased mortality risk. A potential explanation for this differential pattern in risk may be due in part to an earlier onset of overweight and obesity, which would increase the risk of developing obesity-related complications, such as diabetes (23), hypertension (24,25), coronary heart disease (26), and cancer (27,28). Previous research has shown that the duration of overweight and obesity is positively related with the development of metabolic complications (29,30). Further, Abdullah et al. (30) report that the duration of obesity is a risk factor for type 2 diabetes, independent of the level of obesity. Therefore, it can be speculated that individuals with weight gain in early or mid-adulthood may have an earlier onset of overweight and obesity, hence increasing their duration of obesity, and elevating their risk for deleterious health outcomes. These findings indicate that weight gain during this time period is of particular importance for describing mortality risk.

Examination of the health consequences of weight change in late adulthood may be more difficult due to changes in body composition as a result of menopause. For example, the reduction in body weight in older women may be due to decreases in lean mass that may mask concomitant increases in abdominal fat mass (31). High fat mass, and particularly low muscle mass are associated with various deleterious health outcomes in older adults (32,33). Thus, it is possible to speculate that reductions in body weight are reflecting undesirable changes in body composition as a result of menopause, which may lead to increased mortality risk. Moreover, there is an increased potential for these individuals to have underlying diseases, such as cancer and diabetes, which are associated with weight loss as part of the disease process or treatment (1,34,35). For instance, cancer cachexia is a syndrome often experienced by patients with some common forms of cancer, which is characterized by weight loss in the form of muscle loss with or without the loss of fat mass (34,36). Moreover, overweight and obese persons with type 2 diabetes are commonly recommended to lose weight as part of their treatment regime (1). It has been demonstrated that intentional weight loss in overweight individuals with diabetes is associated with significant reductions in mortality risk (35). In our study we adjusted for these diseases in order to lessen their potential confounding effects on weight changes. After adjustments our results still indicated that weight loss during later adult life was associated with increased mortality risk, which is also in agreement with other studies examining weight loss in older adults (13,15,37). However, due to the lack of information on the intentionality of weight loss we cannot rule out the possibility of underlying diseases having an effect on this relationship, and indeed, previous research has indicated the potential importance of taking this into account (15). Nevertheless, these results suggest that weight loss during late adulthood may not decrease mortality risk.

The strengths and limitations of our study warrant mention. Participants in the study were recruited through their own voluntary contact with the clinical centers, and are subject to self-selection biases. Accuracy of weight at ages 18, 35, and 50 years may also be an issue as it was based on retrospective self-report data. However, Stevens et al. (38) report that although the correlations between self-reported weight and actual measured weight decrease with elapsed time (correlation coefficients: 0.98 for current weight, 0.94 for weight 4-years ago, and 0.82 for 28-year recall), they remain quite high overall. Nevertheless, some participants were recalling weight from as long as 50+ years ago, and there remains a lack of information on the validity of self-reported weight beyond 28 years. In spite of this, a strength of the WHI recruitment strategy is that subjects were recruited at various baseline ages, resulting in a cohort of women who were born over a span of 4 decades, thus minimizing the impact of external events from particular time-periods (19). Furthermore, due to the nature of the retrospective self-reported weight data in this study, it was not possible to determine the intentionality of weight loss, and elimination of individuals with pre-existing diseases at baseline was not possible because it would compromise the stability of our regression models.

Despite these limitations, our study has several strengths. To our knowledge, this is the first study that has examined the influence of overall weight changes across adulthood as compared to weight changes during specific age points with mortality risk. By using this methodology we demonstrated a novel component (the timing of weight change), which should be considered in future studies examining the associations between WgtC, WgtV, and mortality risk. Furthermore, our study was conducted in a very large prospective cohort of women recruited from across the United States with up to 9 years of follow-up data. This further adds to the body of literature on the association between WgtC, WgtV, and mortality risk in a cohort that is more prone to weight changes. Additional information is required to complete our understanding of how WgtC and WgtV contribute to deleterious health outcomes in women, and whether these associations differ in men. Particular effort should be made in future studies to examine the intentionality of weight change during adulthood in conjunction with the timing of the change.

In conclusion, merely examining WgtC and WgtV by weight changes across adulthood without taking into consideration the timing of the change may not be adequate to fully describe mortality risk. Thus, future studies investigating WgtC and WgtV should consider the timing of weight change, as it may be an important component in describing mortality risk.

ACKNOWLEDGEMENT

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References

The authors would like to express our greatest appreciation to the study participants and the WHI Investigators. The Women's Health Initiative Observational Study (WHI-OS) is conducted and supported by the National Heart, Lung and Blood Institute (NHLBI) in collaboration with the WHI-OS Study Investigators. This Manuscript was prepared using a limited access dataset obtained from the NHLBI and does not necessarily reflect the opinions or views of the WHI-OS or the NHLBI.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. ACKNOWLEDGEMENT
  8. DISCLOSURE
  9. References