Study Claims No Harm from Weight Cycling Part 2 - Included Studies

This is the Weight and Healthcare newsletter. If you like what you are reading, please consider subscribing and/or sharing!

We are back discussing the article “The Physiological Effects of Weight-Cycling: A Review of Current Evidence” published January 3, 2025. In part 1 we looked at the authors and the introduction, in part 2 we’ll be looking at the studies they included.

They included 23 articles and we’ll take a look at them in some detail. I’ve included the information they they provided in their synopsis in italics and indented. You can skip those parts (which contain weight stigma) and still get the gist of the article. I’ve also included the dates, though they were not included in the synopsis. I think that might be they were trying to avoid questions about their description of “current evidence” being applied to research much of which is from the previous millennium, but that’s pure speculation on my part. As a reminder, this study ignored the physical and psychological harms that have been linked to weight cycling and, instead only considering something “harmful” if it impeded attempts to be thinner.

I do want to note that I did not do deep dives into each of these studies, predominantly because the biggest issues are right there on the surface.

Content note - this piece, including my explanations, contains repeated discussions of weight loss/disordered eating behaviors so if that type of content is difficult for you to read you might want skip this one.

Too Long Didn’t Read Version

It’s a collection of 23 studies that are some combination of:

• Small sample size

• Short duration

• Didn’t actually include higher-weight people

• Weight cycling was not a realistic/real-world scenario

• Analyses not adjusted for confounders (meaning that something else may have caused the effect/lack of effect seen)

There was no consistent definition of weight cycling among the studies.

There did not appear to be any trans or non-binary representation, many of the studies included only cis-women, and in studies where I could find data on race/ethnicity there was often underrepresentation of People of Color which is particularly concerning given the racist nature of BMI/weight stigma/pathologizing higher-weight people*

The study authors made their claims about “overwhelming evidence” based on the number of studies that didn’t find harm (based on their very narrow definition harm) without regard to the sample sizes or demographic representation of those studies.

Okay, let’s get into this:

Van Dale and Saris, 1989
20 ob*se female weight-cyclers aged 20–45 y.
Weight-loss intervention, investigating effects of weight-cycling on weight reduction,
RMR (IC), and body composition (UWW). The intervention was over a 14-week
period with measurements taken at baseline, after week 5, and after week 14. Subjects
were divided into groups on the basis dieting history: frequent dieters (yo-yo, Y) and
non-dieters (non-yo-yo, NY), and also on the basis of BMI into a diet (D) and a diet
and exercise (DE) group, which were further split into three intervention groups: diet
and exercise yo-yo (DE-Y), diet and exercise non-yo-yo (DE-NY), and diet non-yo-yo
(D-NY). The intervention consisted of dietary restrictions (840 kcal/d) and exercise
of 4 h/week with 60% VO2 max for the diet and exercise groups. Analyses were not adjusted for confounders.

So, 20 cis women and a 14 week intervention. Spoiler alert: Small numbers of subjects and short time frames is going to be a bit of a theme here. Also “Analyses were not adjusted for confounders” means that other variables may have impacted or caused any findings.

Melby et al., 1990
Normal weight weight-cycling wrestlers (n12)= with a mean age of 19.4 years and non-wrestling controls (n13)= of similar weight and body composition with a mean age of 20.6 years.

Cohort study measuring RMR (IC) and body composition (UWW) before, during, and after a 6-month wrestling season, a total period of 7 months. Last measure was 5–6 weeks post-season, with weight stabilized for at least 3 weeks. Controls were measured three times over an average period of 4 months. Frequency of previous weight-cycles was self-reported. RMR analysis was adjusted for LBM; body composition analyses were not adjusted for confounders.

A study of collegiate wrestlers. Seriously? This is extrapolatable to cis male college athletes who are intentionally losing about 10 pounds for a sports season, but I’m not sure how applicable it is to any other population. Also, again, small sample and short study duration and body composition analyses were not adjusted for confounders.

Jebb et al., 1991

11 ob*se women, mean age 44 years.
Randomized interventional study. Subjects were recruited from a group of over 120 people who replied to two public advertisements. All subjects were ob*se and had a self-reported history of weight-cycling. The intervention consisted of three weight-cycles of 2 weeks duration with restricted dieting (1861 kJ/day), followed by 4 weeks of no restricted eating. Measurements of RMR (IC) and body composition (UWW, SFT) were taken at baseline, at the end of each diet and free-living period, and at the end of the intervention. Analyses were not adjusted for confounders.

This is a very small sample and an unrealistic premise - three cycles of two weeks each on a diet and then a month not dieting? That is ….not how most weight cycling occurs.

Manore et al., 1991

Non-ob*se female weight-cyclers (n=11) and controls (n=12) matched in height, body weight and composition, activity level, and contraceptive use.
Cross-sectional study measuring body composition (UWW) and RMR (IC) in weight-cycling and weight-stable subjects. Weight-cyclers were recruited based on self-reported weight-cycling history from the preceding 3 years.

They recruited a total of 23 relatively thin women, asked them to remember if they had weight-cycled over the last three years, then compared their current body composition and resting metabolic rate. That does not have the ring of rigorous science. Given that our authors conclusions are around recommending weight cycling to higher-weight (or, as they call them, overw*ight and ob*se” people,) studies on populations that don’t include those people seem not particularly apt.

Prentice et al., 1992
Gambian child-bearing women (n443) and Caucasian British population(n481)
A review comparing data from two previous studies to access associations with weight-cycling and LBM. The study compares populations from the UK versus Gambia, where the British population represent “non cyclers” and the Gambians are unintentional weight-cyclers, due to seasonal food availability. Details of statistical analyses were not reported.

I really don’t have words for this. To be clear, the Gambian women were “weight cycling” because of a lack of food (the study refers to them as “a rural African population that undergoes profound weight cycling caused by an annual hungry season.” Meanwhile, the British women had undergone three cycles of voluntary very low calorie diets which (as expected) predominantly ended up with weight cycling. The issues and number of confounding variables here are absolutely staggering.

Wadden et al. , 1992
Ob*se women (n=50) with a mean age of 40.11± year.
Cross-sectional study. Subjects were drawn from a total of 76 participating in a
6-month weight-reduction intervention. Baseline measurements of body composition
UWW)+(DXA and RMR (IC) were used to investigate associations with weight-
cycling history. Analyses were not adjusted for confounders.

I realize I sound like a broken record: Small sample, short duration, self-reporting and we’re back to “Analyses were not adjusted for confounders.”

McCargar et al., 1993
Normal weight female weight-cyclers (n=9) and control subjects (n=9) matched for age, height, weight, lean body mass, and exercise habits.
Cohort study. Weight patterns, RMR (IC), and body composition (SFT) were assessed
to determine associations weight-cycling. Measurements done at baseline and at
6 and 12 months. The control group had their first test 1 month after the weight-
cyclers, to diminish the likelihood of seasonal variations. Changes in weight, diet, and
exercise were monitored throughout a 1-year period. Analyses were not adjusted for confounders.

18 thin people monitored for a year? The term whoop-de-doo comes to mind. Remember that their claim is that higher-weight people should be told to repeatedly try to lose weight. How does this study help prove that? Not to mention another chorus of “Analyses were not adjusted for confounders.” It would only be less useful if it was a study of, gosh I don’t know, college wrestlers or something.

Schmidt et al., 1993
Normal weight weight-cycling men wrestlers (n6),= with a mean age of 18.8 years
Observational 1 year extension of a previous study by Melby et al.. Measurements
of RMR (IC) and body composition UWW)+(SFT two times during a 1 year
period, pre- and post-wrestling season. RMR analysis was adjusted for LBM; body
composition analyses were not adjusted for confounders.

These are six (6) of the collegiate wrestlers from the Melby study above, but studied for a year. Still problematic in all the same ways.

Van Der Kooy et al., 1993
32 ob*se subjects (17 M; 15 F) aged 25–51 years.
Weight-loss intervention study, measuring changes in body composition (DXA) and
fat distribution (MRI) with weight loss and regain. The subjects were a subsample
of subjects recruited for a dietary-interventional trial including a period of energy
restriction. The weight-loss intervention consisted of a 4.2 MJ/day energy deficit
diet (42% carbs, 25% protein, and 33% fat) for 13 weeks, after the intervention
no counseling was provided. Measurements were performed before and after the
intervention and at follow-up on average 67 weeks after the weight loss. Analyses
were not adjusted for confounders.

So 32 subjects, an average of 67 weeks after a 13 week weight loss intervention. (A nice reminder that weight regain is incredibly common.) Another reminder is that they are only looking at body size measurements, not all the other potential harms. And join me in another rousing chorus of “Analyses were not adjusted for confounders.”

Wadden et al., 1996
Obese women (n=12) with a mean age of 38.38.4± years.
Cross-sectional study. A one-day follow-up. Subjects who had previously been treated
for ob*sity were recruited 2–3 years after weight loss to determine if they had
regained their lost weight. Baseline measurements of weight, body composition
(DXA), and RMR (IC) were compared with values measured at return to baseline,
on average 29.5±3 months after weight loss. Analyses were not adjusted for Confounders.

Here we have a good reminder of the weight loss industry’s various branding attempts. “Treated for ob*sity” just means subjected to a weight loss intervention. Here again we have a grand total of 12 subjects and, say it with me, analyses were not adjusted for confounders.

Hammer et al., 1998
Premenopausal ob*se women (n=44) with at least 30% body fat.
Cross-sectional study, with subjects recruited from one of two previous diet
interventions by the same investigators. Follow-up measurements of body fat
percentage (UWW) were performed at either 6 or 12 months after the previous
interventions. Women were divided into 3 groups: successful maintainer (n14,=
25%<regained of their weight loss), moderately successful (n11,= regained
25–75% of their weight loss), regainers (n19,= 75%>regained of their weight loss).
Comparisons were adjusted for baseline body weight.

First, I have to point out that manipulation here. The dieters were “successful maintainers” if they hadn’t gained back more than 25% of the weight they lost at either 6 or 12 months, were considered “moderately successful” if they didn’t gain back more than 75% of the weight they lost in 6-12 months, and were only considered regainers if they regained 75% or more of the weight in 6-12 months. There is no reason to believe that their weight regain had leveled off at this point or they they wouldn’t all end up regaining all of the weight and possibly more.

Imagine if you heard a news report that a marathoner took the lead and successfully maintained it through the end of the race. Would you think that was a reasonable thing to say as long as she didn’t finish later than 25th place? Or would you assume that if she maintained first place that she ended the race in 1st place? This is one of the ways that words mean different things in weight loss research.

Oh, and they measured the body fat percentage of 44 cis women, that’s it.

Kajioka et al., 2002
Normal weight women (n=5) with no weight-cycling history, aged 22–34 years.
Non-randomized interventional study with sequential phases. 2 weight-cycles, which
consist of 30 days of energy restriction 1200(< kcal/d), followed by 14 days of
free-living (no restrictions). Physical activity was not allowed at any time-point.
Measurements of body weight and composition (BOD-POD) and RMR (IC) were
taken at baseline, at the end of each diet and free-living period, and at 180 days after
baseline measurements. Analyses were adjusted for the change in weight.

First of all, n=5 is not a typo - the entire study was 5 relatively thin women who dieted for 30 days, then took 14 days off (during which they were not allowed to exercise) and they did this cycle twice, then their body composition measurements were taken at the end of each cycling and at 180 days. Again, this is not how weight loss interventions typically work. It’s not just about whether or not someone weight cycled, it’s about how long they were intentionally restricting food as well as other variables. In good news, this is the first study from this millennium.

Leibel et al., 1995
Obese women (n8)= with a mean age of 9±29 years.
Interventional study investigating effects of changes in body weight, RMR (IC), and
body composition (UWW). The intervention consisted of a 10% weight gain from
baseline (6–10 weeks) and a return to baseline with subjects restricted to the formula.
The subjects were restricted to a liquid formula during weight loss period and
maintenance, and their calorie intake of the formula was adjusted until the weight was
stable for 14 days before measurements were taken. Analyses were not adjusted for confounders.

I read this summary and couldn’t believe they could have gotten IRB approval. So I looked up the study (using the data and link provided by these authors) and I’m not sure that the summary above is completely clear. Leibel et al had 18 “ob*se” subjects and 23 subject who had “never been ob*se” who were variously subjected to “over” feeding (they call it “maximally tolerated amount of self-selected foods (generally 5000 to 8000 kcal per day”)) and semi-starvation and/or weight maintenance on a liquid diet (40 percent fat [corn oil], 45 percent carbohydrate [glucose polymer], and 15 percent protein [casein hydrolysate]) supplemented with 5.0 g of iodized sodium chloride, 1.9 g of potassium ions as a potassium salt, and 2.5 g of calcium carbonate per day, 1 mg of folic acid twice weekly, and 36 mg of ferrous iron every other day.)

The authors of the study that I am writing about in this piece are just pulling out 8 of the subjects from the above study.

I still don’t know how in the world they got ethical approval to do this, but obviously there are multiple reasons that this study is not representative of anything resembling normal weight cycling. Oh, by the way - analyses were not adjusted for confounders.

Graci et al., 2004
459 ob*se subjects (340 F; 119 M), age 19–65 years
Cross-sectional study, assessing the relationship between weight-cycling and
cardiovascular risk factors, as well as body composition (BIA). Fat mass was
measured at fasting state. Weight and dieting history were collected by interviewing
the subjects, and 7 parameters were used to determine history of weight-cycling.
Analyses were not adjusted for confounders.

This is a larger study at least but, once again, they are depending on people’s recollection of their history of weight cycling.

Strychar et al., 2009
Postmenopausal ob*se women (n- 121)
Subjects in this cross-sectional study were recruited from a 6-month weight-loss intervention 3–6 years preceding. Baseline and follow-up measurements of body composition (DXA) and RMR (IC).

So this is a slightly larger subject group, though pretty specific in that it is only post-menopausal cis women. One thing I keep remembering is how studies like this demonstrate the prevalence of weight cycling.

Anastasiou et al., 2010
Two groups of normal weight women aged between 18 and 35 years: a high-body-fat group (HBF, n=15) with>30% body fat and a low-body- fat group (LBF, n=17) with≤30% body fat.
Cross-sectional study measuring body composition (DXA). Subjects were divided into groups by age and waist-to-hip ratio. Participants completed the 25-item weight- cycling questionnaire developed by Brunner and Brownell including the 26-item Eating Attitudes Test (EAT-26). Comparisons between groups were not adjusted for any confounders.

This study is just odd - they took 32 cis women who were “normal weight” some with higher body fat than others and looked at their self-reported history of weight cycling. Also, you’ll never guess but “Comparisons between groups were not adjusted for any confounders.”

Lee et al., 2010

147 weight-cyclers of normal weight (49% M; 51% F) aged 70–79 years and controls matched by race and gender (n=147). Cohort study comparing changes in body composition (DXA) in older weight-cyclers with gender- and race-matched weight-stable individuals over a 1-year weight-cycling period. Weight-cycling subjects underwent a single weight loss followed by a regain during the first year of the study and were then followed for a 2-year follow-up period and reevaluated for changes after four and five years. Comparisons (separately in males and females) at baseline were not adjusted for confounders, whereas the 2-year changes were adjusted for age, race, baseline body composition, and weight change

This was a population of 147 “normal weight” senior citizens (ages 70-79) which severely limits extrapolation on multiple axes. Plus, there is a significant amount of evidence that attempting weight loss in that population is ill advised and leads to earlier mortality.

Yoo et al., 2010
Overweight subjects (n109),= divided into weight-cycling (n40= F; 7 M, mean age 41.4 years) and non-weight-cycling (n54= F; 8 M, mean age 44.9 years) groups.
Cross-sectional study with subjects recruited from a 12-week diet intervention program, considering only those who finished the program and completed a self-administrated questionnaire at baseline and after finishing the program. Based on the baseline analysis, subjects were divided into weight-cycling and non-weight-cycling groups. Body composition (BIA) was measured at baseline and after the 12-week intervention. Analyses were adjusted for age.

The explanation suggests several things about this study:

1: The weight loss intervention only lasted 12 weeks

2: Some people dropped out before 12 weeks

3: at least 40 people weight cycled over those 12 weeks

Cereda et al., 2011
914 overweight and ob*se subjects (605 F; 309 M) with a mean age of 51.1.2±14 years.
Open-label cross-sectional study measuring BMI. Covariates in regression models
included sex, age, menopausal status, smoking, education, occupation, physical
activity, and presence of diseases.

This is one of the larger, more recent (relatively,) and more representative samples. And it found that “A history of WC [weight cycling] appears related to body weight excess and abdominal fat accumulation.” I’m not saying there is anything wrong with being fat* or becoming fatter, but this is what the authors of the study we are talking about are defining as “harm.”

One of the things that they did in the summary (that we talked about in part 1,) was claim that an “overwhelming” amount of the evidence found no association between weight cycling and their (very narrow) definition of harm. They based this claim on the number of the 23 included studies that found an association, but didn’t say anything about sample sizes of those studies. This is a possible method of, intentionally or unintentionally, manipulating the evidence to suggest a larger effect (or lack thereof) than actually exists.

Here’s a simple example. Let’s say I’m writing an article to see if an intervention has harmful side effects. I include 5 studies - 4 of them had 10 people each and they find no harm from the intervention I’m investigating. The fifth study includes 1,000 people and it DOES find harm. If I said 4 out of 5 studies didn’t find harm, thus “an overwhelming amount of the evidence” found that there was no harm, that would be incredibly misleading since it would mean that I was claiming that the findings for 40 people were overwhelmingly more accurate than the findings for 1,000 people.

To credibly interpret findings in research like this, we would typically want to take into account population size and diversity, scientific rigor, study length, or other important variables. That’s why credible science doesn’t usually just count up the studies when making claims about effects.

Bosy-Westphal et al., 2013
Overweight and ob*se subjects (n=47: 36 F; 11 M), aged 24–45 years. Not defined as weight-cyclers. Interventional study with subjects drawn from a previous 13-week low-calorie diet intervention. Subjects were divided into two groups based on their weight loss and regain results: weight regainers (n21= F, 6 M, 30%≥regained of their weight loss) and weight stable (n15= F, 5 M, weight 20%±change of their weight loss). The intervention consisted of a 3±13-week low-calorie self-selected diet (800–1000 kcal/ day) with weekly check-ups. Measurements of RMR (IC) and body composition (DXA, MRI, and BOD-POD) were taken at baseline, end of intervention, and at a 6-month follow-up. Analyses were not adjusted for confounders.

This one is just 47 people over 6 months. The weight regainer group was on a 13 week diet. Note that the people they are considering “stable” could actually have regained up to 19.9% of the weight they lost within 6 months of losing it (and be on a trajectory to regain it all or more), but as long as they didn’t hit 20% in 6 months this study called them weight “stable”. Imagine if you asked if you could put your priceless antique on a shelf in my house and I said “absolutely it’s completely stable.” How surprised would you be when you put that keepsake on the shelf and it tilted 19.9 degrees?

Fothergill et al., 2016
14 ob*se subjects (6 M; 8 F), mean age 41.3 years.
Cross-sectional follow-up study with participants from “The biggest loser” weight loss
competition. Follow-up was 6 years after beginning of the competition. The aim of
the study was to measure long-term changes in RMR (IC) and body composition
(DXA) by comparing follow-up measurements to baseline and end of weight loss at the end of the competition (30 weeks). Analyses were not adjusted for confounders.

This is the Biggest Loser study. Here again, this is a small study (only 14 subjects) and all of them were subjected to the extreme physical and emotional (mis) treatment that characterized that horrid reality show so, though they did find that weight cycling had a long-term negative impact, I’m not sure I would consider this any kind of representative sample.

El Ghoch et al., 2017
38 obese subjects (14 M; 24 F) with a mean age of 54.7 ± 13.2 years
Cross-sectional study investigating effects of one weight-cycle on RMR (IC) and
body composition (BIA). Participants were selected from a cohort of 1350 patients
consecutively admitted to a rehabilitative residential treatment program for s*v*re
ob*sity and were considered eligible if they had achieved a weight loss of ≥ 5% after
a prior treatment and were readmitted to the same facility with a regain of ≥ 30%.
Measurements taken at baseline (admission at 1st residential treatment) and follow-up
(admission at 2nd residential treatment). Analyses were adjusted for sex.

Another study showing that weight loss doesn’t work! Another small sample size of 38 people who went to an in-patient weight loss program and then were readmitted after they regained weight. It is not clear that the intervention they would have been subjected to can be extrapolated to people who haven’t been subjected to in-patient weight loss treatment.

Halali et al., 2022
2684 obese subjects (80% women). Cross-sectional study measuring BMI. Subjects were drawn from a previous study in which Finnish adults participated in the 1-year randomized controlled weight-loss trial. Subjects were divided into groups based on their history of trying to lose weight: (1) no; (2) no, but tried to keep my weight stable; (3) 1–2 times; (4) 3 or more; and (5) continuously. BMI analysis was adjusted for age and sex (analyses of other outcomes adjusted for age, sex, and BMI).

This is the largest and most recent study included. It concluded “Multiple attempts to lose weight may unfavourably affect T2D risk factors as well as lifestyle intervention outcomes.”

So this isn’t a great collection of research and the study that is, by far, the largest (in one case almost 537 times larger) found harm though it is harm that they excluded from their definition. Making these findings even less reliable is the fact that there was nothing approaching a common or consistent definition of weight cycling among the studies they included. Definitions range from “Body weight change exceeding 5% of the initial body weight within the previous 2 years” to “Subjects that had a history of weight-cycling during the previous 3.5 years, and weight-cycled > 4.5 kg at least 10 times during the 1st year and at least 5 times during the 2nd year of the study.”

So that’s the research upon which their claims are based. In the final part we’ll look at their analysis and conclusions.

Did you find this post helpful? You can subscribe for free to get future posts delivered direct to your inbox, or choose a paid subscription to support the newsletter (and the work that goes into it!) and get special benefits! Click the Subscribe button below for details:

Liked the piece? Share the piece!

More research
The Research Post

More resources
The Resource Post

*Note on language: I use “fat” as a neutral descriptor as used by the fat activist community, I use “ob*se” and “overw*ight” to acknowledge that these are terms that were created to medicalize and pathologize fat bodies, with roots in racism and specifically anti-Blackness. Please read Sabrina Strings’ Fearing the Black Body – the Racial Origins of Fat Phobia and Da’Shaun Harrison’