OBJECTIVE: To assess the prevalence of clinically significant weight loss among women and whether this is associated with smaller long-term weight gains.
DESIGN: Six-year follow-up of young and middle-aged women in the Nurses' Health Study II.
SUBJECTS: A total of 47 515 women who did not report a pregnancy, or a diagnosis of cancer or cardiovascular disease any time between 1989 and 1995.
MEASUREMENTS: Self-reported weights in 1989, 1991, 1993 and 1995, dietary intake, physical activity, inactivity, history of weight cycling and smoking.
RESULTS: Between 1989 and 1991, 9% of the women lost ≥5% of their 1989 weight (6% lost 5–9.9% and 3% lost ≥10%). The proportion who lost ≥10% of their weight increased with category of body mass index (BMI, kg/m2) from 0.4% among women with a BMI <22 to 9% among women with a BMI ≥30 in 1989. Women who lost ≥5% of their weight between 1989 and 1991 gained more weight between 1991 and 1995 than their peers and the difference increased across categories of BMI in 1989. However, due to their large weight losses, women who lost ≥5% of their weight between 1989 and 1991 overall gained less weight than their peers between 1989 and 1995 (P<0.001). Moreover, women who engaged in 5 or more hours per week of vigorous physical activity gained approximately 0.5 kg less than their inactive peers (P<0.001).
CONCLUSION: Although most women who lost a clinically significant amount of weight regained most of it, they gained less weight over the entire 6 y period than their peers.
Although approximately 40% of adult women are trying to lose weight,1 currently there are few effective treatments to reduce weight and maintain the loss. Therefore intentional weight losses are rarely sustained.2,3 However, little is known about weight change patterns following a large weight loss, other than the fact that few women are able to maintain the weight loss for extended periods. There have been few investigations of the correlates of successful long-term weight maintenance or weight loss maintenance.
Smoking and body weight status have both been associated with weight change. Women who smoke gain less weight than their non-smoking peers,4,5 and those who quit smoking are likely to experience a temporary weight gain.6 In addition, overweight women are likely to gain, or regain, more weight than their leaner peers. The results of the associations between dietary intake, physical activity, and weight change are less consistent. Parker et al7 observed that total energy intake, but no specific nutrients or food group, was associated with weight gain, whereas Klesges et al8 and Lissner et al 9 found that both total energy intake and fat intake were predictive of weight gain. However, Lissner et al observed that the relationship of fat intake to weight gain was of only marginal significance after adjusting for energy intake. In contrast, Colditz et al10 did not find any dietary factors to be predictive of weight change over 8 y among 31 940 women in the Nurses' Health Study.
Although physical activity is frequently recommended for maintaining weight loss, there is a paucity of data directly assessing the role of activity in preventing weight gain or regain in non-clinic based settings. Kahn et al5 observed that walking four or more hours a week was protective against weight gain over a 10 y period. In addition, among 629 women enrolled in the national Weight Control Registry, a registry of adults who have been successful at long-term maintenance of weight loss, weekly physical activity levels were high. However, since the participants self-selected themselves based on the fact that they had been able to maintain their weight loss, it is unclear whether physical activity is independently associated with prevention of weight regain. We are unaware of reports based on data collected prospectively with regular repeated assessments of weight change patterns among adult women. The purpose of this analysis is to assess the prevalence of clinically significant weight loss among a large prospective cohort of adult women and to determine whether these women overall gain more weight over time than their peers. A secondary aim is to assess factors associated with weight change over time.
The Nurses' Health Study II (NHS II) was established in 1989 and consists of 116 671 female nurses, age 25–43 y in 1989, who responded to a mailed questionnaire about their medical history and lifestyle and health behaviors. Follow-up questionnaires were sent to participants in 1991, 1993 and 1995. Additional details have been reported previously.11
Assessment of risk factors
Height, weight and recalled weight at age 18 were ascertained in 1989 and current weight was assessed on each follow-up questionnaire (ie, 1991, 1993 and 1995). The validity of self-reported weight measurements were examined among 140 women, 40–65 y of age, participating in the Nurses' Health Study.12 Self-reported and measured weights were highly correlated (Pearson correlations of 0.97; mean difference between measured and self-reported weights=1.5 kg). The validity of recalled weight at age 18 was ascertained in a sample of 118 participants of the Nurses' Health Study II.13 Body weight at age 18 was only slightly underestimated (mean difference=−1.4 kg) and the correlation between recalled and measured weight at age 18 was high (r=0.87).
Clinically significant weight loss
Women were classified as having had a clinically significant weight loss if their reported weight in 1991 was at least 5% lower than the reported weight in 1989. Women who lost 5–9.9% of their 1989 weight were classified as having a small clinically significant weight loss; whereas, women who lost ≥10% of their 1989 weight were classified as having a large clinically significant weight loss.
Leisure-time physical activity
Recreational physical activity was assessed with eight activity-specific questions (walking or hiking, jogging, running, bicycling, calisthenics/aerobics/aerobic dance/rowing machine, tennis/squash/racquetball, lap swimming or other aerobic recreation). This questionnaire has been validated in a sample of NHS II participants.14 For each item we inquired about average time per week engaged in the activity during the past year. Responses to the 1989 and 1991 questionnaires were averaged to give an estimate of long-term recreational activity level. Total hours per week of activity was estimated by summing across the eight activities. Vigorous activity was the sum of the following six activities: jogging, running, bicycling, calisthenics/aerobics/aerobic dance/rowing machine, tennis/squash/racquetball, lap swimming or other aerobic recreation.
In 1989 and 1991 participants were asked to report the average amount of time per week they spent sitting at home and work. These two one item questions, with nine response categories ranging from zero to more than 90 h per week, were summed together and used as a proxy measure of inactivity. Responses to the 1989 and 1991 questionnaires were averaged to give an estimate of long-term inactivity.
Women were classified as never, past, or current smokers based on their responses to the 1989–1995 questionnaires. Women who reported on each questionnaire (1989, 1991, 1993 and 1995) that they did not smoke were classified as never smokers. Women who reported that they were a current smoker in 1989, but no longer smoked in 1991 or women who reported smoking in 1998 or 1991, but did not smoke in 1993 were classified as past smokers. To be classified as a current smoker, a woman had to report that she smoked on each of the questionnaires. Because many women gain weight soon after they quit smoking, women who quit smoking between 1993 and 1995 were excluded from the analysis since they could not be accurately classified as current or past smokers.
A previously validated, 116-item semi-quantitative food frequency questionnaire15,16,17 was used to assess diet, including alcohol intake, in 1990–1991. The questionnaire asks participants how often on average they consume a specific portion size of each of the 116 foods listed. Response categories range from never to more than six per day. Nutrient intake was computed by multiplying the frequency of consumption by nutrient content, estimated from standard food composition sources. For each of the micro and macronutrients, the frequency of consumption was multiplied by the nutrient content and summed over all the food items.15 Among a sample of Nurses' Health Study I participants, alcohol intake assessed by the food frequency questionnaire was highly correlated with intake assessed from diet records (Spearman r=0.90).18
The questions on weight losses were specifically designed to address the long-term health consequences of intentional weight loss. They were developed after extensive discussion among investigators from the Nurses' Health Study, Centers for Disease Control and University of Minnesota. The questions asked on the 1993 questionnaire included the number of intentional weight loss episodes of varying magnitude (5–l9 pounds (2.3–4.1 kg), 10–19 pounds (4.5–8.6 kg), 20–49 pounds (9.1–22.2 kg), and 50+ pounds (≥22.7 kg)) over the past 4 y and between 18 and 30 y. The information on intentional weight losses between age 18 and 30 was used to classify women as non-, mild or severe weight cyclers. The question, ‘Between the ages of 18 and 30, how many times did you lose each of the following amounts of weight on purpose?’ The responses were 0, 1–2, 3–4, 5–6 and 7 or more times for each of the magnitudes of weight loss (5–9 pounds (2.3–4.1 kg), 10–19 pounds (4.5–8.6 kg), 20–49 pounds (9.1–22.2 kg), and 50+ pounds (≥22.7 kg)). To be consistent with the magnitude of the weight loss required by Field et al,19 French et al,20 and Williamson et al 21 in their studies of the relation between intentional weight loss and disease, we required that a woman report intentionally losing 20 or more pounds (≥9.1 kg) to be considered a weight cycler. To ensure that the cyclers were women who had repeatedly lost weight, we required that women intentionally lost weight three or more times between ages 18 and 30 to be classified as severe weight cyclers. Women who had intentionally lost 10 or more pounds (≥4.5 kg) three or more times between ages 18 and 30, but did not meet the criteria for severe weight cycling, were classified as mild weight cyclers. For example, a woman who lost 20 or more pounds (≥9.1 kg) twice and 15 pounds (6.8 kg) once would be classified as a mild weight cycler. To be classified as a severe weight cycler she would have to have lost 20 or more pounds (≥9.1 kg) on each of at least three intentional weight losses. Women who did not meet the criteria described above for mild or severe weight cycling were classified as non-weight cyclers. The weight change variable was not used to define weight cycler status.
Two definitions of weight maintenance were used. The more stringent definition required that a woman gain no more than 5 pounds (2.3 kg) from her 1991 weight (ie, achieved weight after loss). The other definition of maintenance required that a woman gain no more than 5% of her 1991 weight. Using these definitions, women who continued to lose weight were classified as successful weight losers.
Weight change was defined as the difference (in kg) between weights reported on two questionnaires. Two time frames were used to delineate weight change. One was the difference in self-reported weights in 1989 and 1995 (ie, over the entire 6 y period), whereas the other weight change outcome was restricted to the period from 1991 to 1995 (ie, the 4 y after which women had lost a clinically significant amount of weight). The purpose of focusing on weight change over the four year period following the large weight loss (ie, 1991 to 1995) is to determine whether women who lose a clinically significant amount of weight gain more weight during the subsequent time period than their peers who did not lose large amounts of weight. However, to assess whether women who lose large amounts of weight, weigh less in the long run, even if they regain some of their weight soon after their weight loss, it is necessary to include the entire 6 y period from 1989 to 1995.
Sample for analysis
For this analysis we excluded women who did not complete the 1991 (n=8226), 1993 (n=5093), or 1995 (n=4134) questionnaires, did not report dietary intake (n=10 299), had experienced a pregnancy between 1988 and 1995 (n=31 793), had a history of cardiovascular disease (n=709) or cancer (n=1576), did not provide information on weight in 1989, 1991, 1993 or 1995 (n=5007), reported an implausible height (n=229), did not provide information on smoking status in 1989, 1991 or 1993 (n=251), did not provide information on physical activity in 1989 or 1991 (n=272), reported more than 40 h per week of vigorous activity (n=5), did not complete questions on intentional weight losses between age 18 and 30 (n=150), or were determined to be an outlier, using Rosner's method,22 in terms of their weight change 1989 to 1995 or 1991 to 1995 (n=347). In addition, women who did not meet the criteria for mild or severe weight cycling, but reported 1–2 50 pounds (≥22.7 kg) losses (n=209) were excluded from the analysis since they had reported a major intentional weight loss, but did not fit into any prespecified category. Moreover, women who quit smoking between 1993 and 1995 were excluded because they could not be accurately classified as current or past smokers (n=859). Therefore 47 515 women remained for analysis.
The analysis focused on two outcomes, weight change and weight loss maintenance. In the latter analyses the sample was limited to women who had a clinically significant weight loss between 1989 and 1991, whereas in the weight change analysis all women were included, but the primary contrast was between women who had a clinically significant weight loss (5–9.9% or ≥10%) and their peers who had lost <5% of their weight between 1989 and 1991.
Linear regression was used to assess whether women who had a clinically significant weight loss between 1989 and 1991 had different weight change patterns than their peers, independent of body mass index (BMI) at age 18, weight change from age 18 to 1989, smoking (never, past, current), age, dietary intake (as a percentage of total calories or in deciles, of intake), alcohol intake (in deciles), hours per week of sitting (including time at home and work), hours per week engaged in physical activity, and a history of weight cycling between the ages of 18 and 30. In addition, to assess whether the impact of a clinically significant weight loss on subsequent weight change varied by baseline BMI, interactions terms were included as covariates. Among the subgroup of women who had a clinically significant weight loss between 1989 and 1991, logistic regression was used to determine whether any of these factors, or the amount of weight lost between 1989 and 1991, predicted successful weight maintenance from 1991 to 1995.
During the 2 y period from 1989 to 1991, 2590 (5.5%) women lost 5–9.9% of their 1989 weight and 1326 (2.8%) women lost at least 10% of their 1989 weight. The proportion of women who lost ≥5% of their baseline weight increased with category of BMI (kg/m2) from 3% (5–9.9% weight loss) and 0.4% (≥10% weight loss) among women with a BMI <22 to 9% among women with a BMI ≥30 in 1989 (Figure 1). Over the next 4 y, 1991–1995, approximately 50% of these women regained all of the weight they had lost. Among the women who lost ≥10% of their 1989 weight, the percentage who regained all of their large weight loss between 1989 to 1991 decreased across baseline categories of BMI from 71% among the women with a BMI <22 to 54% among the women with a BMI≥30 in 1989 (Figure 2).
Less than 10% of the women who had a large clinically significant weight loss between 1989 and 1991 were able to successfully maintain their weight loss (Figure 2). The proportion was greater among the women with smaller weight losses (ie 5–9.9%); however, it still less than 20% (Figure 3). Regardless of the definition of weight maintenance (ie gained less than 5 pounds (2.3 kg) or gained less than 5% of 1991 weight), baseline category of BMI was inversely associated with successful weight loss maintenance. For example, among the women who had a large clinically significant weight loss, the percentage who gained less than 5% from their 1991 weight decreased from 15% among the leanest women (BMI<22) to 4% among the heaviest women (BMI≥30).
The women who lost greater than or equal to 10% of their weight between 1989 and 1991 gained more weight between 1991 and 1995 than their peers who did not lose weight (Table 1). The difference between the groups increased across categories of BMI in 1989. For example, the difference between women who did not lose weight between 1989 and 1991 and those who lost ≥10% of their weight increased from 3.2 kg (ie 5.0−1.8) among women with a BMI <22 to 11.8 kg among women with a BMI≥30. However, due to their large weight losses, women who lost ≥10% of their weight between 1989 and 1991 had a net weight gain less than their peers between l989 and 1995 (P=0.001). For example, among women with a BMI≥30, those who lost 10% of their weight between 1989 and 1991 gained 6.8 kg less than their peers who had not lost weight.
Age, smoking, BMI at age 18, and weight gain from age 18 to 1989 were associated with subsequent weight change and therefore were included in all multivariate linear and logistic regression models. After adjusting for these factors, women who lost at least 5% of their weight between 1989 and 1991, with the exception of women who had a BMI<22 in 1989 and lost ≥10% of their weight over the next 2 y, gained more weight between 1991 and 1995 than their peers (Table 2). Although the leanest women (ie those who had a BMI<22) gained on average 1.2 kg less than their peers who had not had a clinically significant weight loss, the women in higher categories of BMI gained more than their peers. Each one category increase in baseline BMI was associated with gaining an additional 3 kg compared to women who did not lose a clinically significant amount of weight (Table 2). For example, women with a baseline BMI between 22 and 24.9 who had lost ≥10% of their weight between 1989 and 1991, had an average net weight gain of 1.8 kg (ie − 1.23 kg+2.99 kg) more than their peers between 1991 and 1995. However, when considering the entire course of the study (1989–1995), women in all categories of baseline BMI who had a clinically significant weight loss between 1989 and 1995 (5–9.9% or 10%) gained less weight than their peers (Table 3 ).
BMI at age 18 and weight change during early adult life were predictive of weight gain (Tables 2 and 3). Each one unit difference in BMI at age 18 was associated with gaining an additional 0.32 kg between 1989 and 1995. In other words, a woman with a BMI of 25 at age 18 gained approximately 1.6 kg more than her peer who had a BMI of 20. However, the association was slightly attenuated when a history of weight cycling between ages 18 and 30 was entered into the statistical model. Both mild and severe weight cyclers gained more weight than their peers between 1989 and 1995, as well as between 1991 and 1995. Mild weight cyclers gained approximately 1.8 kg more and severe cyclers gained 3.4 kg more between 1989 and 1995 than their peers (Table 3).
Vigorous physical activity was protective against weight gain. Women who engaged in five or more hours per week of vigorous activity gained approximately 0.5 kg less than their inactive peers between 1989 and 1995 (Table 3). In contrast, total hours of activity per week (including walking) was not associated with weight change (data not shown). Whereas, physical inactivity was associated with weight gain. For each 10 h a week a woman spent sitting at home or work, she gained approximately 0.11 kg more than her less inactive peers.
Dietary intake during the year 1990–1991 was modestly associated with subsequent weight change from 1991 to 1995. Energy intake had a modest positive association with weight gain, whereas, alcohol intake was associated with less weight gain. For example, an increase in decile of energy intake was associated with a 0.04 kg weight gain (Table 2). Percentage of calories from fat was not associated with subsequent weight change; however, there was a modest positive association between percent of calories from protein and weight gain from 1991 to 1995.
In further analyses limited to the subset of women who had a clinically significant weight loss between 1989 and 1991, we assessed weight regain. Among these women, the more weight they lost between 1989 and 1991, the more likely they were to maintain the loss between 1991 and 1995. Each 10 pounds weight loss approximately doubled the probability that a woman would maintain her weight loss (odds ratio (OR)=1.95, 95% confidence interval (CI) 1.66–2.29). However, women with a history of weight cycling between the ages of 18 and 30 were less likely to maintain their weight loss from 1991 to 1995. Women who had been severe weight cyclers between ages 18 and 30 were approximately 40% less likely to maintain their weight loss (OR=0.57, 95% CI 0.39–0.85), whereas, those who had been mild weight cyclers were 20% less likely to maintain the loss (OR=0.79, 95% CI 0.64–0.96). There was no association between inactivity or intake of total calories of dietary fat (data not shown), however hours engaged in vigorous activity was a significant predictor of weight loss maintenance 1991–1995. Each hour of vigorous activity increased the likelihood by 7% (OR=1.07, 95% CI 1.01–1.14) of maintaining the weight over the 4 y.
We observed that approximately 9% of the women in the sample lost a clinically significant amount of weight between 1989 and 1991, but few of these women were able to maintain their weight losses for at least 4 y. The percentage of women who had a clinically significant weight loss increased with category of BMI in 1989; however, the proportion of women who were able to maintain successfully the large weight loss was inversely related to category of BMI in 1989. Although most of the women who lost a clinically significant amount of weight regained most of it, because of the substantial weight loss before 1991, they gained less weight over the entire 6 y period than their peers.
Excessive weight and weight gain are major public health concerns in the United States. Obesity is a risk factor for cardiovascular disease23,24 and total mortality.25 Moreover, excessive weight and weight gain increases the risk of developing hypertension26,27 and diabetes,28 and exacerbate many other chronic diseases such as hypertension29,30,31,32 osteoarthritis,33,34 gallstones,35 dyslipidemia and musculoskeletal problems.36,37,38 The severity of the consequences of excess weight and weight gain are the reason that the current weight guidelines advise adults to maintain their current weight or lose weight.39 Despite the recommendation, there has been little population-based research on weight maintenance or weight loss maintenance. However, both observational studies and randomized clinical trials have observed that physical activity is associated with weight loss maintenance.40,41
Our data support the importance of physical activity as an effective means to prevent weight gain. Schoeller et al estimated that an average of 35 min a day of vigorous activity would result in weight maintenance among 32 middle-aged women who had recently lost weight.42 Our results offer support for their finding. We observed that the vigorous activity was protective against weight regain among the subset who had lost a clinically significant amount of weight between 1989 and 1991. In addition, we observed that overall women who engaged in five or more hours per week of vigorous activity gained less weight than their inactive peers. Physical activity also has been observed by other investigators to be predictive of weight loss maintenance, particularly among individuals who have participated in weight loss interventions.40,41 These data offer support for the recommendation that people engage in physical activity to prevent weight regain.
One possible reason that we did not observe an association between total hours of activity and weight change is that moderate activity is measured with more error than vigorous activity. For most adults, the majority of the time they engage in physical activity is spent in moderate intensity activities, such as walking. Thus, total activity is measured with greater error than vigorous activity. Therefore, although we did not observe such an association, it is possible that moderate activity is beneficial in preventing weight gain and regain. In addition to the benefits of activity, we found that hours per week of sitting at home or work (a proxy measure for inactivity) had an independent positive association with weight gain. Thus our data suggest that public health messages should be two-pronged, promoting increasing activity and decreasing inactivity. Not only should women be encouraged to engage in physical activity as a means to control their weight, but also they should be encouraged to reduce the amount of time they are inactive, such as watching television.
In addition, the data suggest that weight maintenance and weight gain prevention efforts should be targeted at young adults since BMI at age 18, weight gain from age 18 to 1989, and history of weight cycling between the ages of 18 and 30 were all independently predictive of adult weight gain. Thus it appears that behaviors or lifestyle factors associated with weight control, or lack thereof, are established by late adolescence. Our results are consistent with those of Pasman et al 43 and French et al,44 who observed that women who diet to lose weight gained, or regained, more weight than their non-dieting peers. Research is needed on how to prevent the development of these behaviors.
Our finding that energy intake is positively associated with weight gain is consistent with the results seen by both Parker et al 7 and Lissner et al.9 However, unlike Lissner or Klesges et al,8 we did not see an association with fat intake. Toubro and Astrup45 observed that among obese patients in a randomized clinical trial a high carbohydrate diet was more predictive than fixed energy intake of maintaining weight after a large weight loss, but many subjects were lost to follow-up and the results were marginally significant in only one of three comparisons. Nevertheless, our finding regarding percentage of calories from protein should be interpreted cautiously. When the dietary data was collected in 1991, the popular belief was that a diet high in carbohydrates was beneficial for weight control and weight loss. Therefore, it is not clear whether a diet with a higher percentage of calories from carbohydrates (as opposed to protein or fat) is in itself helpful for preventing weight gain or whether in this analysis it was serving as a proxy measure for intentionally trying to lose or maintain weight. A limitation of the current study is that we did not include an assessment of efforts, such as changes in dietary intake, used to lose or control weight. Moreover, there are data to suggest that overweight adults underreport their energy intake more than lean adults.46,47 Thus, we cannot rule out the possibility that we underestimated the true relationship between energy intake and weight change. In the analyses we conducted we adjusted for predictors of weight change, several of which also are predictors of underreporting of intake (age, smoking status and BMI). The inclusion of these confounders hopefully minimized the bias.
Another limitation of the current study is that some of the weight losses of at least 5% of the women's 1989 weight may have been unintentional. We do not know how many women might have lost the weight unintentionally, however, our best estimate is from the question we included on the 1997 Nurses' Health Study II questionnaire. In 1997 less than 13% of the women reported an unintentional weight loss in the past 2 y and fewer than 20% of those women reported that their unintentional weight losses were at least 6.8 kg (15 pounds). These results suggest that the group of weight losses of 5–9.9% might contain relatively more women with unintentional losses than the group comprised of women with losses of at least 10% of their weight. By including some unintentional weight losses with losses that were intentional we may have slightly underestimated the true relationships between weight change, diet, physical activity, inactivity and past weight change history. Therefore, our estimates should be considered conservative.
We did not include women who had a pregnancy during the time period of study, thus it is not clear whether these results are generalizable to mothers with very young children. The relationship between pregnancy and long-term weight gain is much more complicated and an appropriate investigation of the topic would need to take into consideration weight gain during pregnancy, breastfeeding (which will change caloric needs), changes in diet and activity, and timing and spacing of pregnancies (ie other pregnancies during the follow-up). Therefore we felt that it would not be prudent to include women with pregnancies during the follow-up in the analysis.
In conclusion, our findings suggest that, although few women can completely maintain weight losses, women should not be discouraged from attempting to lose weight since we observed that women who had a clinically significant weight loss gained less weight than their peers over the 6 y period of study. Moreover, clinicians should counsel their patients about the benefits of increasing activity, as well as decreasing inactivity. For patients who are resistant to adopting vigorous activity, it may be advisable to recommend decreasing inactivity as a first step in adopting a more active lifestyle that will help them prevent weight gain or regain.
Dr Field was partially supported by the Boston Obesity Nutrition Research Center (DK 42600). Additional funding was provided by a research grant (CA50385-09) from the National Institutes of Health, a cooperative agreement with the Centers for Disease Control and Prevention (S040-11/15), and a First Independent Research Support and Training Award (R29) grant (HL57871-01) from the National Institutes of Health.