Abstract
Objective
This study aimed to identify major themes of a large cohort experiencing long‐term weight‐loss maintenance who answered open‐ended questions about weight‐loss triggers, current motivations, strategies, and experiences.
Methods
Machine learning and topic modeling were used to analyze responses to six open‐ended questions among 6,139 WW International, Inc., (formerly Weight Watchers) members with weight‐loss maintenance; inclusion criteria included ≥9.1‐kg loss with weight‐loss maintenance for ≥1 year.
Results
Participants (mean age = 53.6 years; 94.3% White; mean BMI = 27.8 kg/m2) had lost 24.5 kg and maintained the loss for 3.4 years. Descriptions of factors triggering weight loss coalesced into five topics: medical status, appearance, mobility, social prompts, and change needed. Factors currently motivating weight‐loss maintenance yielded two topics: looking back at experiences at higher weight and health/appearance concerns. Advice for others to succeed in weight‐loss maintenance coalesced on two recommendations: perseverance in the face of setbacks and consistency in tracking. Rewards for weight management included improved confidence, pain, mobility, fitness, body image, medical status, and affect. Two thematic negative consequences were clothing costs and sagging skin.
Conclusions
Future weight‐maintenance research should include more diverse populations and investigate weight‐loss maintenance as a journey with highs and lows, perseverance in the face of setbacks, sustained tracking, and making changes in medical status more salient during the weight‐maintenance journey.
Study Importance.
What is already known?
-
►
Registries of weight‐loss maintainers have analyzed data from closed‐ended, researcher‐determined questions to describe the factors that allow some people to succeed at weight‐loss maintenance.
-
►
Weight‐loss maintainers have reported eating a low‐calorie diet, engaging in high levels of physical activity, frequent self‐monitoring, problem solving, setting daily intake goals, limiting sitting time, and keeping low‐calorie foods accessible.
What does this study add?
-
►
This is the first study, to our knowledge, to use machine learning and topic modeling to analyze the written responses of more than 6,000 weight‐loss maintainers who described their motivations, strategies, struggles, and successes with long‐term weight management.
-
►
Weight‐loss maintainers in WW International, Inc., (formerly Weight Watchers) advised perseverance in the face of setbacks, consistency in food tracking, and looking back at experiences of life at higher weight and described improved health and appearance as motivating factors in their weight‐loss journeys.
How might these results change the direction of research?
-
►
Findings may lead researchers to study interventions that promote perseverance in the face of setbacks, encourage tracking of intake, and make changes in medical status more salient during the weight‐loss journey.
INTRODUCTION
Modest weight loss can reduce long‐term risk of cardiometabolic disease, but these improvements are attenuated with weight regain, which affects most individuals. Biological, behavioral, psychological, sociocultural, and environmental pressures promote weight regain. Nevertheless, about 20% of individuals who have lost weight in the US population are able to keep it off long term and experience ongoing improvements in quality of life and health status (1, 2, 3).
Research of weight‐loss maintainers from the WW International, Inc., (formerly Weight Watchers) Success Registry (WWSR), the National Weight Control Registry, and the German and Portuguese Weight Control Registries have characterized factors that allow some people to succeed at long‐term weight‐loss maintenance (4, 5, 6, 7). Among the modifiable factors are eating a lower‐calorie diet, engaging in high levels of physical activity, frequent self‐monitoring, problem solving, setting daily intake goals, limiting sitting time, and keeping low‐calorie foods accessible (6, 8, 9, 10). Psychological strategies include cognitive restraint (11) and frequent practice of healthy coping skills such as “thinking about past successes” and “remaining positive in the face of weight regain” (9).
These large‐scale registry studies have included validated questionnaires, and findings have led to novel advancements in intervention research (12, 13, 14). However, reliance on closed‐ended, researcher‐determined questions may fail to capture the rich array of thoughts, feelings, values, behaviors, and beliefs that surround the experience of long‐term weight‐loss maintenance. Qualitative studies with focus groups and in‐depth interviews have uncovered new perspectives. For example, a 2015 meta‐synthesis of qualitative research of weight‐loss maintainers characterized “sources of tension” (e.g., old habits and impulses, pleasure, discomfort in new body image) that may disrupt weight‐maintenance success and identified several strategies to manage this tension (e.g., self‐regulation skills) or reduce it (e.g., developing automaticity, changing beliefs and self‐concept) (15). However, this meta‐synthesis and other qualitative research in the area have relied on small sample sizes and employed different methodologies, analysis techniques, and definitions of successful weight‐loss maintenance. Large‐scale registry studies of weight‐loss maintainers have generally shied away from open‐ended questions owing, in part, to the labor‐intensity of analyzing textual responses from thousands of participants. However, technological advances now allow for analysis of textual responses from large data sets.
The purpose of this study was to identify major themes in the responses of a large cohort of long‐term weight‐loss maintainers in a commercial weight‐management program who answered open‐ended questions about their motivations, strategies, and experiences. Machine learning and topic modeling were used to analyze responses of more than 6,000 long‐term weight‐loss maintainers.
METHODS
This is a mixed method, cross‐sectional investigation of weight‐loss maintainers in the WWSR. The WWSR is an observational study of individuals who lost weight in WW and were successful at long‐term (≥1 year) maintenance of substantial (≥9.1 kg) weight loss. WW is a widely available behavioral weight‐management and wellness program that has been studied extensively (16, 17, 18, 19). Prospective weight‐loss maintainers were recruited through an email sent by WW to members who had reported a loss in WW of ≥9.1 kg for at least 1 year. Interested individuals were referred to the study website hosted by California Polytechnic State University (Cal Poly), San Luis Obispo, for online screening, consent, and enrollment. Eligibility was based on self‐reported weight, height, weight change, and duration of weight loss. To be eligible for enrollment, individuals self‐reported age ≥ 18 years and maintenance of ≥9.1 kg loss from WW entry for ≥1 year. Study procedures were approved by the Cal Poly Institutional Review Board, and all participants provided informed consent electronically via Research Electronic Data Capture (REDCap; Vanderbilt University, Nashville, Tennessee).
Measures and open‐ended questions
All measures were administered online via REDCap immediately after consent. Closed‐ended questions asked participants standard questions on demographics (age, education level, income, marital status, race/ethnicity, and gender) and details about weight history (weight‐loss duration, age of onset of overweight, and maximum lifetime weight), as well as current weight and height (7, 20).
A series of open‐ended questions were designed to capture the factors that prompted their successful weight loss, the factors that continued to motivate them and promote success, and the consequences of weight‐loss maintenance. All questions are listed in Table 1. There were no character limits set on responses. Participants were not required to respond to the open‐ended questions.
TABLE 1.
Weight‐loss triggers |
|
Motivation |
|
Advice |
|
Changes |
|
Other |
|
Of the 7,419 enrolled in the WW Success Registry, 82.7% (n = 6,139) responded to at least one open‐ended question, 74.6% responded to at least three, 64.2% responded to at least five, and 12.9% responded to all six of the open‐ended questions.
Statistical analysis
Descriptive statistics were used to describe participant characteristics, and independent t tests and χ2 tests were used to compare the characteristics of WWSR participants who responded versus participants who did not respond to the open‐ended questions.
A machine‐learning, natural‐language‐processing approach with Latent Dirichlet Allocation (LDA) was used to analyze the open‐ended questions. LDA is an unsupervised (i.e., data‐determined) topic‐modeling method (21, 22). The method is similar to a cluster analysis. LDA assumes that the collection of responses to each question contain probability distributions of latent topics, and topics contain probability distributions of words. Word frequency and co‐occurrence of words are used to cluster words and phrase patterns together and to form underlying topics. LDA assigns probabilities (i.e., word weights) of each word appearing in each underlying topic.
Prior to LDA analysis in the current study, text was segmented into words (i.e., tokenization), and words with little informational value were removed (i.e., of, which, the, such). Also, an auto‐stemming process was implemented to arrive at a word root. For inclusion in the analysis, no minimum word length was set. The topic‐modeling implementation parameter (eta) was set to 0.01 to minimize overlap between topics for each question, and three models were trained for each question: a two‐topic model, a three‐topic model, and a five‐topic model. From these quantitatively extracted topic models, the research team determined the optimal number and names of topics following an iterative process (21) based on the following: 1) topic coherence (i.e., how well the words in each topic related conceptually); (22) 2) the top 20 most frequent words within each topic, devaluing words with low frequencies (<50 incidences in five‐topic models and <100 incidences in two‐topic models) and those that were in the open‐ended question itself (e.g., “weight,” “motivate”); 3) consideration of word weights (i.e., the probability the corresponding word belonged to that topic) (22); and 4) a review of a random selection of ≥100 participant responses containing the words associated with each topic. For presentation in the paper, a subset of the team (SP, JR, and SMN) selected representative quotes for each topic that were narrowed down and lightly edited for grammatical errors (Tables 3, 4, 5, 6, 7) (23). R (R Foundation, Vienna, Austria) and SPSS Statistics version 25.0.0 (IBM Corp., Armonk, New York) were used in analyses of descriptive statistics, χ2, and t tests. Python (Python Software Foundation, Wilmington Delaware) was used for the unsupervised, machine‐learning analysis.
TABLE 3.
Topics (% responses) a Words (frequency/ weighted probabilities) b | Representative quotes |
---|---|
1. Medical (15.3%) |
|
Health (2,111/0.071) | |
Live (93/0.060) | |
Diabetes (194/0.054) | |
Heart (115/0.036) | |
2. Appearance (25.2%) |
|
Clothes (599/0.070) | |
Fit (373/0.050) | |
Look (483/0.039) | |
Feel (346/0.032) | |
3. Mobility (16.4%) |
|
Walk (80/0.44) | |
Stairs (62/0.007) | |
Knee (128/0.006) | |
4. Social prompts (18.0%) |
|
WW (389/0.055) | |
Friend (186/0.027) | |
Doctor (193/0.025) | |
Husband (140/0.016) | |
Daughter (138/0.016) | |
Child (228/0.015) | |
5. Change needed (26.1%) |
|
Tired (662/0.070) | |
Picture (243/0.021) | |
Fat (230/0.014) | |
Change (147/0.016) | |
Age (124/0.033) |
% responses indicate the proportion of the responses reflected in the topic.
Weighted probabilities indicate the extent to which a word is uniquely associated with a topic; although words within each topic were not analyzed jointly, they clustered within responses and shared a unique association with the same topic. Results are from analysis with stemming. This is not an exhaustive list of the 20 words extracted for each topic. Shown are the most frequent (>50) and highly weighted words in each topic.
TABLE 4.
Topics (% responses) a Words [frequency/ weighted probabilities] b | Representative quotes |
---|---|
1. Looking back (41.5%) |
|
Back (593/0.054) | |
Goal (219/0.024) | |
Gain (465/0.021) | |
Work (228/0.019) | |
Year (153/0.015) | |
Time (132/0.015) | |
Hard (119/0.011) | |
2. Health and appearance (58.5%) |
|
Health (1,953/0.075) | |
Clothes (1,068/0.067) | |
Fit (489/0.039) | |
Feel (483/0.034) | |
Enjoy (151/0.011) |
Abbreviation: CPAP, continuous positive airway pressure.
% responses indicate the proportion of the responses reflected in the topic.
Weighted probabilities indicate the extent to which a word is uniquely associated with a topic; although words within each topic were not analyzed jointly, they clustered in responses and shared a unique association with the same topic. Results are from analysis with stemming. This is not an exhaustive list of the 20 words extracted for each topic. Shown are the most frequent (>100) and highly weighted words in each topic.
TABLE 5.
Topics (% responses) a Words (frequency/ weighted probabilities) b | Representative quotes |
---|---|
Perseverance (54.5%) |
|
Give (855/0.048) | |
Day (680/0.045) | |
Time (681/0.043) | |
Work (430/0.029) | |
Meet (410/0.023) | |
Goal (306/0.021) | |
Stick (336/0.018) | |
Small (209/0.011) | |
Journey (310/0.012) | |
Tracking and lifestyle (45.5%) |
|
Track (748/0.057) | |
Eat (417/0.050) | |
Change (420/0.035) | |
Lifestyle (375/0.030) | |
Healthy (227/0.017) | |
Slow (191/0.014) | |
Habit (149/0.014) |
% responses indicate proportion of the responses reflected in the topic.
Weighted probabilities indicate the extent to which a word is uniquely associated with a topic; although words within each topic were not analyzed jointly, they clustered in responses and shared a unique association with the same topic. Results are from analysis with stemming. This is not an exhaustive list of the 20 words extracted for each topic. Shown are the most frequent (>100) and highly weighted words in each topic.
TABLE 6.
Topics (% responses) a Words (frequency/weighted probabilities) b | Representative quotes |
---|---|
Confidence (19.4%) |
|
Confidence (939/0.236) | |
Health (839/0.182) | |
Self (680/0.140) | |
Energy (335/0.051) | |
Happy (181/0.029) | |
Esteem (108/0.029) | |
Reduced pain (23.0%) |
|
Pain (115/0.051) | |
Life (200/0.033) | |
Exercise (66/0.031) | |
Walk (113/0.031) | |
Back (64/0.027) | |
Time (80/0.023) | |
Fitness and body image (17.4%) |
|
Active (122/0.072) | |
Clothing (184/0.068) | |
Level (194/0.067) | |
Fit (184/0.063) | |
Physical (250/0.053) | |
Ability (126/0.042) | |
Energy (335/0.042) | |
Move (64/0.022) | |
Medical status (21.7%) |
|
Blood (149/0.077) | |
Healthier (122/0.068) | |
Medications (149/0.064) | |
Cholesterol (53/0.026) | |
Longer (128/0.032) | |
Diabetes (64/0.022) | |
Positive affect (18.5%) |
|
Feel (178/0.055) | |
Eat (156/0.052) | |
Body (108/0.036) | |
Appear (66/0.020) | |
Comfort (82/0.026) | |
Control (58/0.016) | |
Life (200/0.012) |
Abbreviation: CPAP, continuous positive airway pressure.
% responses indicate the proportion of the responses reflected in the topic.
Weighted probabilities indicate the extent to which a word is uniquely associated with a topic; although words within each topic were not analyzed jointly, they clustered in responses and shared a unique association with the same topic. Results are from analysis with stemming. This is not an exhaustive list of the 20 words extracted for each topic. Shown are the most frequent (>50) and highly weighted words in each topic.
TABLE 7.
Topics (% responses) a Words (frequency/weighted probabilities) b | Representative quotes |
---|---|
Clothing and criticism (38.3%) |
|
Clothing (582/0.073) | |
People (359/0.039) | |
Eat (358/0.039) | |
Buy (310/0.038) | |
Money (172/0.022) | |
Friend (154/0.020) | |
Time (132/0.015) | |
Skin and effort required (61.7%) |
|
Skin (516/0.063) | |
Food (183/0.018) | |
Back (159/0.018) | |
Hard (139/0.016) | |
Saggy (142/0.010) |
% responses indicate the proportion of the responses reflected in the topic.
Weighted probabilities indicate the extent to which a word is uniquely associated with a topic; although words within each topic were not analyzed jointly, they clustered in responses and shared a unique association with the same topic. Results are from analysis with stemming. This is not an exhaustive list of the 20 words extracted for each topic. Shown are the most frequent (>100) and highly weighted words in each topic.
RESULTS
Of the 7,419 enrolled, 82.7% (n = 6,139) responded to at least one open‐ended question and thus were included in this study. Participant characteristics are displayed in Table 2. In examining the characteristics of participants who responded versus those who did not respond to the open‐ended questions, they differed on magnitude of initial weight loss since starting WW (mean [SD], 24.5 [12.5] kg vs. 23.5 [11.9] kg, p = 0.007), current WW membership (88.3% vs. 82.0%, p = 0.001), being married (74.1% vs. 67.9%, p = 0.005), and being White (94.3% vs. 33.2%, p = 0.001), but they did not significantly differ on age, current BMI, gender, duration of weight maintenance, income, education, employment, or other demographic variables (data not shown). Weight‐loss maintainers had lost, on average, 24.3 kg from their WW starting weight and maintained weight loss for an average of 3.4 years. They were, on average, 53.6 (12.6) years old; 91.9% identified as female; 94.3% identified as White; 71.2% were married; 65.7% were employed; 64.5% had an annual income more than $75,000; and 88.2% had a college education or more.
TABLE 2.
Age at WW enrollment, mean (SD), y | 49.4 (12.8) |
Age at study enrollment, mean (SD), y | 53.6 (12.6) |
Female (%) | 91.9 |
Currently in WW (%) | 87.9% |
Lifetime maximum weight, mean (SD), kg | 105.5 (23.1) |
Weight at start of WW, mean (SD), kg | 101.3 (21.5) |
Lowest weight, mean (SD), kg | 71.1 (15.5) |
Current weight, mean (SD), kg | 77.0 (17.0) |
Weight loss since WW start, mean (SD), kg | 24.5 (12.5) |
Duration of ≥9.1‐kg loss from WW start weight, mean (SD), y | 3.40 (3.79) |
Weight lost from maximum weight, mean (SD), kg | 28.5 (15.0) |
Current BMI, mean (SD), kg/m2 | 27.8 (5.56) |
BMI categories | |
Obesity (%) | 23.7 |
Overweight (%) | 43.6 |
Normal weight (%) | 32.6 |
Underweight (%) | 0.0 |
Income (total in family per year) | |
<$25,000 (%) | 4.9 |
$25,000‐75,000 (%) | 30.6 |
≥$75,000 (%) | 64.5 |
Race/ethnicity | |
White (%) | 94.3 |
Black (%) | 2.8 |
Hispanic (%) | 4.0 |
Employed (%) | 65.7 |
College education or more (%) | 88.2 |
Married (%) | 71.2 |
Sources of motivation
When asked about the factors that prompted successful weight loss, responses clustered into five topics (Table 3): topic 1) “Medical Triggers,” reflecting a high number of words related to medical status (health, live, diabetes, heart, risk, insulin); topic 2) “Appearance,” containing a high number of words related to concerns over looks (clothing, fit, look, feel); topic 3) “Mobility,” reflecting concerns related to bodily movement (walk, stairs, knee); topic 4) “Social Prompts,” describing the role others played prompting their weight loss (WW, friend, doctor, husband, daughter, child); and topic 5) “Change Needed,” reflecting feelings of hitting a maximum threshold of tiredness (tired); this topic also included terms that were more scattered among different subjects (e.g., picture, fat, age, change).
Participants were also asked to describe what currently motivated them, and responses clustered around two main topics (Table 4): topic 1) “Looking Back,” centering around a desire to avoid negative experiences of the past given all the time and work that went into achieving their goal (back, goal, work, year, time, hard); and topic 2): “Health and Appearance,” which focused on the desire to maintain health and feel good in clothes (health, clothing, fit, feel, enjoy).
Strategies for success
Participants were asked about the one piece of advice that they would give to help someone else succeed at long‐term weight loss. Responses clustered around two main topics, and they are shown in Table 5. “Perseverance” reflected an approach that encouraged never giving up, taking it day by day, using meetings to reset after difficult weeks, and embracing the journey and long‐term goal (give, day, time, work, meet, goal, stick, small, journey). Topic 2 was labeled “Tracking and Lifestyle,” as weight‐loss maintainers described tracking food intake as an essential skill within a healthy lifestyle (track, eat, change, lifestyle, healthy, slow, habit).
Consequences of weight‐loss maintenance
Participants were asked about the most important thing that has changed as a result of weight loss. Responses are in Table 6, and they clustered around five topics: topic 1) “Confidence,” reflecting improved self‐confidence, esteem, and happiness (confidence, health, self, energy, happy, esteem); topic 2) “Reduced Pain,” including descriptions of less pain doing exercise and activities of daily life (pain, life, exercise, walk, back, time); topic 3) “Fitness and Body Image,” which described an active lifestyle and fitting into clothes (active, clothing, level, fit, physical, ability, energy, move); topic 4) “Medical Status,” which included feeling healthier and improvements in blood pressure, cholesterol, and diabetes and a longer life; and topic 5) “Positive Affect,” which included terms about feeling at ease and more comfortable in mind and body (feel, comfort, body, appear, life) but also included other words that made this topic less coherent (eat, control).
When asked about negative consequences of successful weight loss, responses centered on two topics (Table 7). Topic 1 was “Clothing and Criticism,” reflecting words about the money spent buying new clothes and unexpected criticism from other people; this cluster also included other, more scattered themes (clothing, people, eat, buy, money, friend, time). The second topic was labeled “Skin and Effort Required,” reflecting words about excess skin and the hard work involved in weight‐loss maintenance (skin, back, hard, body, sagging).
Participants were also given the opportunity to describe other factors affecting their weight history. Only a minority of the sample (n = 1,010; 13.6%) responded. Words were mostly scattered, but a commonality was the word WW (or the former program name, Weight Watchers), describing comments on the WW program itself (e.g., “One thing I wasn't able to express is how important my personal support has been ‐ maybe not always from my spouse or family members, but from other WW members and friends”). Another participant described, “The support found in meetings and relationships with other WW members has been instrumental in keeping me accountable to maintaining the weight loss.” A second and related word was “time,” describing the number of times participants had “been down this road” and joined and rejoined WW (e.g., “I had joined WW so many times, but this time I was determined to stick with it.”).
DISCUSSION
This is the first study, to our knowledge, to use topic modeling to understand experiences of more than 6,000 weight‐loss maintainers in WW who wrote about their motivations, strategies, struggles, and successes with long‐term weight control. Weight‐loss maintainers described obesity‐related memories, health, and appearance concerns as motivating both the initiation and sustainment of long‐term weight loss. Perseverance in the face of setbacks and tracking food intake over time were recommended to others seeking similar weight‐loss success. Profound rewards for their weight‐management efforts were described, including major improvements in confidence, pain, mobility, body image, and mental and physical health. Some negative consequences were clothing costs, sagging skin, and unexpected criticism from others. In addition to common experiences, weight‐loss maintainers described unique pathways in the weight‐loss maintenance process.
Appearance concern was identified as a major motivator of both weight loss and maintenance. Participants described initiating successful weight‐loss efforts out of intense feelings of “shame” and “disgust” after looking at themselves in the mirror or in photographs; many described “not knowing who they’d become” and feeling as if they were “ruining” other people’s pictures of events. Participants also described profound embarrassment while shopping for clothes and being unable to fit into clothes, and the ability to fit into clothes was reported by many as the most important thing that changed as a result of weight loss. These findings suggest that appearance concerns may shape not only weight loss but weight maintenance. The persistence of appearance‐based concerns may reflect a distressing sociocultural environment that ties appearance with self‐concept, stigmatizes and discriminates against people with obesity, and promotes an arbitrary definition of beauty based on a Western thin ideal (24, 25). Today, 70% of the US population is living with overweight or obesity; however, many weight‐loss maintainers described initiating weight‐loss efforts out of a pressing desire to “become normal.” Systems and social‐ecological‐based approaches to weight management may hold promise for combatting the prevailing sociocultural climate of weight‐based stigma and discrimination and resulting self‐esteem (26, 27).
Medical health was identified as another powerful theme motivating successful weight loss and maintenance. Prior research on the role of medical factors in motivating weight management has been mixed (28, 29). Among long‐term weight‐loss maintainers in the National Weight Control Registry, “medical triggers,” such as being told by a physician to lose weight or receiving a medical diagnosis, resulted in greater initial weight losses and better maintenance of weight loss over time than nonmedical triggers for weight loss or no trigger at all (28). By contrast, other studies have suggested that medical triggers may lead to less weight loss (29). One clinical trial (30) compared treatments that emphasized medical/health versus appearance concerns and found that treatments that emphasized appearance resulted in more (~3 kg) 6‐month weight loss and better maintenance at 12 and 18 months (30). It is possible that appearance concerns are more salient motivators for younger individuals and that medical factors become more salient over time as the health threats imposed by obesity mount and medical care for obesity‐related comorbidities becomes needed (31).
A powerful piece of advice that weight‐loss maintainers had for others was clear: persevere in the face of inevitable setbacks. Weight‐loss maintainers depicted their successful weight loss as “a long marathon with individual victories and setbacks” requiring “great perseverance.” When faced with inevitable setbacks, weight‐loss maintainers recommended restarting anew the next day or as soon as possible. Outside of the weight‐control literature, research has suggested that perseverance can be promoted by modifying perceptions of effort (32). In educational settings, “grit” has been considered teachable and defined as “perseverance to accomplish long‐term or higher‐order goals in the face of challenges and setbacks” (33). Future research in weight management should prioritize the inclusion of validated measures of perseverance and explore novel approaches to promote it during weight maintenance.
The other major theme that emerged was the importance of tracking, which aligns squarely with prior research (34). Self‐monitoring has long been considered a core component of standard behavioral obesity treatment, and most evidence‐based treatment approaches, including the WW program, encourage tracking. In the current study, the term “lifestyle” was closely tied to tracking, encouraging people to see tracking and other health behaviors as part of necessary lifestyle changes. One long‐term weight‐loss maintainer advised that “you have to get up every day and make a choice to track and eat right. It is going to be difficult and there will be days that you will fall, but you can get back up and keep moving forward. This is a lifestyle change, not a diet.”
The statements describing consequences of weight‐loss maintenance were resoundingly positive; participants described how weight‐loss maintenance had transformed nearly every facet of their lives, including self‐confidence, pain, fitness and mobility, medical status, and affect. For some, these positive improvements served as an ongoing source of motivation: “This is something that is hard fought and hard won and something that will keep me motivated for the future.” Although positive rewards outweighed negative ones, there were some common negative consequences of weight‐loss maintenance, including the expense of having to buy a new wardrobe and body‐image concerns related to having saggy skin.
This study is the first, to our knowledge, to use machine learning and topic modeling to analyze textual responses of over 6,000 long‐term weight‐loss maintainers describing their motivations, strategies, and consequences of weight‐loss maintenance. The use of an unsupervised, machine‐learning approach allowed for efficient analysis and identification of broad themes in a large volume textual data set. Advantages of an unsupervised method is the opportunity for researchers to discover themes that might have been missed through human‐supervised methods. Novel themes that emerged included the role of remembering painful experiences as a motivator and the concept of perseverance. Although powerful themes emerged, there were several other topics that were more scattered that did not easily converge, underscoring a rich diversity of individual experiences and pathways to long‐term successful weight management. One weight‐loss maintainer wrote “Everyone’s journey is a unique education into learning about oneself and what works for your lifestyle. Self‐discovery takes a while, so don't give up!”
This study also has several limitations. Even unsupervised, machine‐learning approaches have subjective elements that can lead to a biased point of view (35). The number and label assignment of quantitatively derived themes in this study were, to some extent, subjective, and research is still emerging on optimal ways to choose the proper number of topics in a model beyond a major iterative approach (36). Because this was an unsupervised approach, the positive and negative valiance of some topics were unclear (e.g., social prompts for weight loss reflected both positive and emotionally hurtful social experiences). We examined term frequency and probability weights but did not examine word vectors (cluster bag of words) or skip grams (in which a word predicts surrounding words). Future research is needed using validated measures of the topics and themes identified in this study, including perseverance, motivation, stigmatization, and health‐related quality of life.
Another major limitation in the current study is that participants were predominantly White, female, married, educated, and with at least midlevel income. Also, within the WWSR cohort, people who chose versus those who did not choose to complete the open‐ended questions were more likely to be White. Therefore, findings from this study may only generalize to a primarily White subset of the population of weight‐loss maintainers. This is a significant study weakness because the prevalence of obesity and related comorbidities is highest in more socially disadvantaged groups, including people of color who have less education and income. It is likely that the prioritized themes might differ based on these and other characteristics that were not examined in the current study. Moreover, the experiences reported among participants in this study might not align with those who could not afford the WW program or those who lost weight via other programs or self‐selected means. Future research is needed to understand the motivations, strategies, and experiences of weight‐loss maintainers from different race/ethnicities and socioeconomic backgrounds and those from different programs, as well as those with varying degrees of weight loss.
In conclusion, this topic analysis suggested that weight‐loss maintainers in WW were commonly motivated by health and appearance concerns and that they experienced profound improvements in these and other domains as a result of weight‐management efforts. Future weight‐maintenance research should include more diverse populations and investigate the importance of promoting perseverance in the face of setbacks, sustained tracking, and making changes in medical status more salient during the weight‐maintenance journey. Societal level interventions that confront a prevailing sociocultural climate in which weight‐based self‐esteem and discrimination persist are needed. Although common themes emerged from the experiences of more than 6,000 weight‐loss maintainers, unique pathways were also described in the lives of those engaged in the ongoing process of weight‐loss maintenance.O
CONFLICT OF INTEREST
SP reports receiving a research grant from WW International, Inc. MIC and GDF are current employees and shareholders of WW International, Inc. The other authors declared no conflict of interest.
AUTHOR CONTRIBUTIONS
SP had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: SP, JR, HG, and GDF; analysis and interpretation of data: HG, JR, SP, and SMN; data collection and management: NA; drafting of the manuscript: JR, SMN, and SP; critical revision of the manuscript for important intellectual content: SP, MIC, and GDF; administrative, technical, or material support: NA; and study supervision: SP.
Phelan S, Roake J, Alarcon N, et al. In their own words: Topic analysis of the motivations and strategies of over 6,000 long‐term weight‐loss maintainers. Obesity (Silver Spring). 2022;30:751–761. doi: 10.1002/oby.23372
Funding information
This research was supported by a grant from WW International, Inc., and student fellowship support from the William and Linda Frost Fund at California Polytechnic State University in San Luis Obispo, California.
REFERENCES
- 1. Wing RR, Phelan S. Long‐term weight loss maintenance. Am J Clin Nutr. 2005;82(suppl 1):222S‐225S. [DOI] [PubMed] [Google Scholar]
- 2. Field AE, Wing RR, Manson JE, Spiegelman DL, Willett WC. Relationship of a large weight loss to long‐term weight change among young and middle‐aged US women. Int J Obes Relat Metab Disord. 2001;25:1113‐1121. [DOI] [PubMed] [Google Scholar]
- 3. Weiss EC, Galuska DA, Khan LK, Gillespie C, Serdula MK. Weight regain in US adults who experienced substantial weight loss, 1999‐2002. Am J Prev Med. 2007;33:34‐40. [DOI] [PubMed] [Google Scholar]
- 4. Santos I, Vieira PN, Silva MN, Sardinha LB, Teixeira PJ. Weight control behaviors of highly successful weight loss maintainers: the Portuguese Weight Control Registry. J Behav Med. 2017;40:366‐371. [DOI] [PubMed] [Google Scholar]
- 5. Neumann M, Holzapfel C, Muller A, Hilbert A, Crosby RD, de Zwaan M. Features and trajectories of eating behavior in weight‐loss maintenance: results from the German Weight Control Registry. Obesity (Silver Spring). 2018;26:1501‐1508. [DOI] [PubMed] [Google Scholar]
- 6. Thomas JG, Bond DS, Phelan S, Hill JO, Wing RR. Weight‐loss maintenance for 10 years in the National Weight Control Registry. Am J Prev Med. 2014;46:17‐23. [DOI] [PubMed] [Google Scholar]
- 7. Klem ML, Wing RR, McGuire MT, Seagle HM, Hill JO. A descriptive study of individuals successful at long‐term maintenance of substantial weight loss. Am J Clin Nutr. 1997;66:239‐246. [DOI] [PubMed] [Google Scholar]
- 8. Roake J, Phelan S, Alarcon N, Keadle SK, Rethorst CD, Foster GD. Sitting time, type, and context among long‐term weight‐loss maintainers. Obesity (Silver Spring). 2021;29:1067‐1073. [DOI] [PubMed] [Google Scholar]
- 9. Phelan S, Halfman T, Pinto AM, Foster GD. Behavioral and psychological strategies of long‐term weight loss maintainers in a widely available weight management program. Obesity (Silver Spring). 2020;28:421‐428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Pascual RW, Phelan S, La Frano MR, Pilolla KD, Griffiths Z, Foster GD. Diet quality and micronutrient intake among long‐term weight loss maintainers. Nutrients. 2019;11:3046. doi: 10.3390/nu11123046 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Teixeira PJ, Carraca EV, Marques MM, et al. Successful behavior change in obesity interventions in adults: a systematic review of self‐regulation mediators. BMC Med. 2015;13:84. doi: 10.1186/s12916-015-0323-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Wing RR, Tate DF, Gorin AA, Raynor HA, Fava JL. A self‐regulation program for maintenance of weight loss. N Engl J Med. 2006;355:1563‐1571. [DOI] [PubMed] [Google Scholar]
- 13. Tate DF, Jeffery RW, Sherwood NE, Wing RR. Long‐term weight losses associated with prescription of higher physical activity goals. Are higher levels of physical activity protective against weight regain? Am J Clin Nutr. 2007;85:954‐959. [DOI] [PubMed] [Google Scholar]
- 14. Gorin AA, Raynor HA, Fava J, et al. Randomized controlled trial of a comprehensive home environment‐focused weight‐loss program for adults. Health Psychol. 2013;32:128‐137. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Greaves C, Poltawski L, Garside R, Briscoe S. Understanding the challenge of weight loss maintenance: a systematic review and synthesis of qualitative research on weight loss maintenance. Health Psychol Rev. 2017;11:145‐163. [DOI] [PubMed] [Google Scholar]
- 16. Jolly K, Lewis A, Beach J, et al. Comparison of range of commercial or primary care led weight reduction programmes with minimal intervention control for weight loss in obesity: Lighten Up randomised controlled trial. BMJ. 2011;343:d6500. doi: 10.1136/bmj.d6500 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Heshka S, Anderson JW, Atkinson RL, et al. Weight loss with self‐help compared with a structured commercial program: a randomized trial. JAMA. 2003;289:1792‐1798. [DOI] [PubMed] [Google Scholar]
- 18. Gudzune KA, Doshi RS, Mehta AK, et al. Efficacy of commercial weight‐loss programs: an updated systematic review. Ann Intern Med. 2015;162:501‐512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Ahern AL, Wheeler GM, Aveyard P, et al. Extended and standard duration weight‐loss programme referrals for adults in primary care (WRAP): a randomised controlled trial. Lancet. 2017;389:2214‐2225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. McGuire MT, Wing RR, Klem ML, Lang W, Hill JO. What predicts weight regain in a group of successful weight losers? J Consult Clin Psychol. 1999;67:177‐185. [DOI] [PubMed] [Google Scholar]
- 21. Zhao WZ, Zou W, Chen JJ. Topic modeling for cluster analysis of large biological and medical datasets. BMC Bioinformatics. 2014;15(suppl 11):S11. doi: 10.1186/1471-2105-15-S11-S11 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Newman D, Lau JH, Grieser K, Baldwin T. Automatic evaluation of topic coherence. Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computional Linguistics. ACM; 2010:100‐108. [Google Scholar]
- 23. Dutta SK, O'Rourke EM. Chapter 13: Open‐ended questions: the role of natural language processing and text analytics. In: Macey WH, Fink AA, eds. Employee Surveys and Sensing: Challenges and Opportunities. Oxford University Press; 2020:202‐218. [Google Scholar]
- 24. Puhl RM, Lessard LM, Himmelstein MS, Foster GD. The roles of experienced and internalized weight stigma in healthcare experiences: perspectives of adults engaged in weight management across six countries. PLoS One. 2021;16:e0251566. doi: 10.1371/journal.pone.0251566 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Frederick DA, Saguy AC, Sandhu G, Mann T. Effects of competing news media frames of weight on antifat stigma, beliefs about weight and support for obesity‐related public policies. Int J Obes (Lond). 2016;40:543‐549. [DOI] [PubMed] [Google Scholar]
- 26. Puhl RM, Suh Y, Li X. Legislating for weight‐based equality: national trends in public support for laws to prohibit weight discrimination. Int J Obes (Lond). 2016;40:1320‐1324. [DOI] [PubMed] [Google Scholar]
- 27. Vo L, Albrecht SS, Kershaw KN. Multilevel interventions to prevent and reduce obesity. Curr Opin Endocr Metab Res. 2019;4:62‐69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Gorin AA, Phelan S, Hill JO, Wing RR. Medical triggers are associated with better short‐ and long‐term weight loss outcomes. Prev Med. 2004;39:612‐616. [DOI] [PubMed] [Google Scholar]
- 29. Borgatti A, Tang ZT, Tan F, Salvy SJ, Dutton G. Predicting program attendance and weight loss in obesity interventions: do triggering events help? J Health Psychol. 2021;26:2056‐2061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Kalarchian MA, Levine MD, Klem ML, Burke LE, Soulakova JN, Marcus MD. Impact of addressing reasons for weight loss on behavioral weight‐control outcome. Am J Prev Med. 2011;40:18‐24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. LaRose JG, Leahey TM, Hill JO, Wing RR. Differences in motivations and weight loss behaviors in young adults and older adults in the National Weight Control Registry. Obesity (Silver Spring). 2013;21:449‐453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Mrazek AJ, Ihm ED, Molden DC, Mrazek MD, Zedelius CM, Schooler JW. Expanding minds: growth mindsets of self‐regulation and the influences on effort and perseverance. J Experiment Soc Psychol. 2018;79:164‐180. [Google Scholar]
- 33. U.S. Department of Education, Office of Educational Technology . Expanding Evidence Approaches for Learning in a Digital World. U.S. Department of Education; 2013. [Google Scholar]
- 34. Butryn ML, Phelan S, Hill JO, Wing RR. Consistent self‐monitoring of weight: a key component of successful weight loss maintenance. Obesity (Silver Spring). 2007;15:3091‐3096. [DOI] [PubMed] [Google Scholar]
- 35. Bazeley P, Jackson K. Qualitative Data Analysis with NVivo. 2nd ed. Sage Publications; 2013. [Google Scholar]
- 36. Zhao WZ, Chen JJ, Perkins R, et al. A heuristic approach to determine an appropriate number of topics in topic modeling. BMC Bioinformatics. 2015;16(suppl 13):S8. doi: 10.1186/1471-2105-16-S13-S8 [DOI] [PMC free article] [PubMed] [Google Scholar]