Deep divefrom Episode 7TrainingFat loss21 October 202622 min

Hard Intervals, Small Trials, and the Fitness That Holds Up

A deep dive into one paper: what was known before, what the researchers did, what the numbers say and what they don't.

The paper: Effects of high-intensity interval training on body composition, cardiorespiratory fitness, and selected lipid outcomes in adult women with overweight or obesity (Wang et al., Frontiers in Physiology, 2026)

MP3

Chapters

Transcript

The script the two AI voices read from. The times are worked out from the script, not heard, so a line can start a few seconds early or late.

Read the whole episode

Cold open

Host 1You heard the trailer — this is the whole story. Welcome to the Hastrup Review of Science — where artificial intelligence reads real science, so you don't have to pretend you understood the paper. Today, fifteen small trials, a few hundred women, and the question of what short, hard bursts of exercise actually do for women living with overweight or obesity.

Host 2The paper is called Effects of high-intensity interval training on body composition, cardiorespiratory fitness, and selected lipid outcomes in adult women with overweight or obesity: a systematic review and meta-analysis. It has seven authors, with Yulong Wang as first author and Ming Li as corresponding author.

Host 1Where are they from, and where did it appear?

Host 2Mostly the School of Physical Education and Sports Science at Fujian Normal University in Fuzhou, plus Anglo-Chinese College in Fuzhou and Huaibei Normal University. It is peer reviewed and was published in Frontiers in Physiology on the twenty-second of September twenty twenty-six.

Host 1And what did we actually read?

Host 2The full text, but our copy cuts off partway through the results, after the fat mass analysis. So for fitness and cholesterol I have the pooled numbers, the subgroup tables and the evidence grading, but not the authors' detailed sensitivity checks or their discussion section, and I will flag where that matters.

Host 1The claim, in two sentences.

Host 2Compared with women who did no structured exercise, women who did interval training ended up with somewhat lower weight and body fat and clearly better aerobic fitness, and with better HDL and LDL cholesterol. But the trials are small, varied and mostly at some risk of bias, and the authors themselves rate the certainty of every result as low or very low.

Why this field exists

Host 1Let's start with the basics. What is interval training, and why did it become its own research field?

Host 2Background, not from this paper. Interval training means repeated bouts of hard effort separated by easier recovery, rather than one long steady session. Athletes have trained this way for most of a century, but it moved into health research in the two-thousands, when studies suggested that much shorter sessions could improve fitness about as much as long moderate ones.

Host 1Why would that matter for people with obesity specifically?

Host 2Because lack of time is one of the most common reasons people give for not exercising, and adherence to long programmes is poor. If twenty-odd minutes of intervals three times a week gets most of the benefit, that is a more realistic prescription. The authors make exactly this argument in their introduction.

Host 1And why look only at women?

Host 2The authors cite global estimates that adult obesity in women rose from eight point eight percent in nineteen ninety to eighteen point five percent in twenty twenty-two. They also argue that estrogen, menopause, conditions like polycystic ovary syndrome and female patterns of fat storage make women's response worth studying on its own, since many earlier trials pooled men and women or studied mainly men.

What was known before

Host 1So what did we already know before this review?

Host 2Background again. Several earlier meta-analyses in mixed adult groups found that interval training reliably raises aerobic fitness, and reduces body fat modestly. When it has been compared head to head with steady moderate exercise, the fat loss has generally come out similar, with intervals mainly saving time.

Host 1And what does exercise alone usually do to body weight?

Host 2Established background: exercise without dietary change typically produces small weight losses, often a couple of kilograms over a few months, because people tend to compensate by eating a bit more or moving less the rest of the day. Its bigger and more reliable effects are on fitness and metabolic health rather than the number on the scale.

Host 1And fitness is measured how?

Host 2With VO two max or VO two peak: the most oxygen your body can use per minute during an all-out test on a treadmill or bike. Background: it is one of the strongest predictors of long-term cardiovascular health and survival, with large observational studies linking each step up in fitness to a lower risk of death.

The open question

Host 1So what was the open question?

Host 2Whether those effects hold up when you restrict the evidence to women with overweight or obesity, and look at several outcomes together instead of one at a time. The authors chose nine: body weight, body mass index, body fat percentage, fat mass, aerobic fitness, and four blood lipids, namely HDL cholesterol, LDL cholesterol, total cholesterol and triglycerides.

Host 1Why separate weight from fat?

Host 2Because weight and body mass index cannot tell fat from muscle. If training builds a little muscle while burning some fat, the scale may barely move even though body composition improved, so fat percentage and fat mass are the more direct measures.

Host 1And what did they say the review was not meant to answer?

Host 2They are explicit that it does not cover blood pressure, insulin resistance, blood sugar control or inflammation. And they say it was not designed to find an optimal interval prescription, only to describe the evidence that exists.

What the paper did

Host 1So walk me through what they did.

Host 2They followed the standard reporting guideline for systematic reviews, called PRISMA, and registered the review in advance in the international registry PROSPERO. They searched four databases, PubMed, Web of Science, Embase and MEDLINE via Ovid, from the start of each database to the twenty-sixth of July twenty twenty-six.

Host 1Anything they did not search?

Host 2Yes, and they say so. They did not search the Cochrane trials register or Scopus, and did not systematically search trial registries or unpublished grey literature, which they list as a limitation that may have missed studies. The search also appears to have been revised during peer review: the paper was received in June, but the search ran to late July, with outcome terms removed so it would catch more trials.

Host 1How did the numbers shake out?

Host 2Three thousand five hundred and seventy-eight records, of which nine hundred and thirty-four were duplicates. Two reviewers independently screened the remaining two thousand six hundred and forty-four, ninety-three went to full-text review, and seventy-eight of those were excluded, thirty-eight because they were not randomised trials. Seven more turned up from reference lists, but five could not be retrieved and the other two were not randomised, so the final count was fifteen trials.

Host 1What qualified a trial?

Host 2Randomised, adult women only or with women's data reported separately, and overweight or obesity defined by the trial itself, usually a body mass index of twenty-five or thirty and above. They also accepted one trial of so-called normal-weight obesity: women with a normal body mass index but body fat above thirty percent.

Host 1And what counted as interval training?

Host 2Repeated vigorous to near-maximal bouts with recovery in between, by running, walking, cycling, circuits or functional exercises. All-out sprint interval training was excluded when the original authors labelled it that way. One trial had both a two-minute and a one-minute protocol, and only the two-minute arm went in, because the one-minute arm counted as sprints.

Host 1And the comparison group?

Host 2This is the most important design choice. The control groups did no structured exercise at all: usual lifestyle, usual care, health education or a placebo. So every result in this paper is interval training versus doing nothing new, not versus walking or steady jogging.

Host 1Any trials with extra ingredients?

Host 2One trial also tested an algae supplement, Chlorella vulgaris. The authors used only the arm that did intervals plus a placebo and the arm that took placebo alone, so the supplement could not muddy the comparison.

Host 1Describe the trials themselves.

Host 2Published between twenty fifteen and twenty twenty-three, with interval groups of eight to thirty women each. They ranged from Chinese college students around twenty years old, to women with polycystic ovary syndrome, to postmenopausal women with metabolic syndrome, breast cancer survivors, women with obesity and high blood pressure, and one trial in women with an average age of sixty-six.

Host 1And the workouts?

Host 2A classic template was four bouts of four minutes at eighty-five to ninety-five percent of maximum heart rate, with three minutes of easier recovery between them. Others used thirty-second runs, circuits of eight movements, or hour-long circuit sessions. Most programmes ran three sessions a week, and they lasted from three weeks up to about eighteen, with one at twenty-four.

Host 1How do you combine results measured in kilograms, percentages and millimoles?

Host 2With a standardised effect size called Hedges' g: the difference between the groups divided by the spread of values within them, with a correction for small samples. Background: by a common rule of thumb, zero point two is small, zero point five is medium and zero point eight is large, though those labels were never meant to be universal.

Host 1And the model?

Host 2Random-effects meta-analysis, which assumes the true effect really differs from trial to trial and estimates the average. Each of the nine outcomes was pooled separately.

Host 1One detail I want to pin down: what exactly was compared?

Host 2Values at the end of training, not change from the start. The authors say change scores were not reported consistently enough to pool, so they compared the two groups' final values and checked starting differences only descriptively. Hold on to that, because it comes back in the caveats.

Host 1How did they judge the quality of the trials?

Host 2With the Cochrane risk-of-bias tool, version two, two reviewers independently. Not one of the fifteen trials came out at low risk overall. Eleven had some concerns, mostly poorly reported randomisation, effort-dependent tests and missing protocols, and four were at high risk, mainly because of large or uneven dropout.

Host 1And the overall certainty?

Host 2They used the GRADE system, which starts randomised evidence at high certainty and then downgrades it. Every outcome lost a level for risk of bias and another for imprecision, and most lost a third for inconsistency between trials.

A word from our sponsor

Host 1Before the results, a word from our sponsor. Do you do four-minute bouts at ninety percent of your maximum heart rate and worry about leaks? Then you are in luck.

Host 1The Hastrup Review of Science is brought to you by menstrual cups. Soft, reusable, and designed to stay in place through every interval, every recovery and every cool-down.

Host 2I wear mine to spin class. It holds firm right through the high-intensity phase.

Host 1Menstrual cups come in several sizes and firmnesses, they are in stock right now, and they are very reasonably priced.

Host 2If you have strong pelvic floor muscles, go for the firmer model. It really stands up to the pressure.

Host 1Our referral link is in the show notes.

Results

Host 1Body weight first.

Host 2Eleven trials, three hundred and thirty-six women. Interval training came out ahead with a Hedges' g of minus zero point three three, confidence interval minus zero point five eight to minus zero point zero eight. That is small, and the trials agreed fairly well, with heterogeneity of about twenty-four percent, which is low.

Host 1In kilograms?

Host 2The paper does not translate it, and stresses that the pooled numbers are standardised rather than in natural units. My own rough back-of-envelope: body weight in these trials varied by around ten to fifteen kilograms between women, so a g of zero point three three is on the order of a few kilograms, and it should be treated as an illustration, not a result.

Host 1Body mass index?

Host 2Eleven trials, three hundred and forty-three women, g of minus zero point five six, interval minus zero point nine five to minus zero point one seven. Here the trials disagreed much more, with heterogeneity of about sixty-eight percent.

Host 1Body fat percentage?

Host 2Ten trials, two hundred and ninety-seven women, g of minus zero point seven six, interval minus one point three five to minus zero point one seven. Heterogeneity was eighty-three percent, which is high: the trials were not telling the same story about how big the effect is.

Host 1And fat mass?

Host 2Eight trials, two hundred and twenty-nine women, g of minus zero point six six, interval minus one point one one to minus zero point two zero, with heterogeneity of sixty-four percent. So all four body measures point the same way, and the fat measures show bigger standardised effects than weight, which fits the idea that some fat was replaced by lean tissue, though the paper does not test that directly.

Host 1Did those results depend on any single trial?

Host 2They checked by dropping each trial in turn, and every body composition result stayed negative and significant. But the size moved. Without one trial of postmenopausal women with metabolic syndrome, by Kazemi and colleagues, the body mass index effect shrank from minus zero point five six to minus zero point four, and heterogeneity fell from about sixty-eight to twenty-five percent.

Host 1And body fat?

Host 2The same pattern with a different trial: the normal-weight obesity study in Chinese college students, by Hu and colleagues. Without it, the body fat effect shrank from minus zero point seven six to minus zero point five two, and heterogeneity fell from eighty-three to forty-four percent. So the direction is stable, but how large the effect looks depends heavily on one or two trials.

Host 1Now fitness.

Host 2This is the clearest result. Twelve trials, three hundred and seventy women, and a g of one point zero nine, with an interval of zero point seven to one point four seven. Even the bottom of that interval is a medium-sized effect, and every subgroup the authors looked at pointed the same way.

Host 1Did the type of exercise matter?

Host 2In this one case the subgroups did differ formally. Running or walking intervals gave a g of one point four one across six trials, cycling zero point four nine across four, and circuits one point two two across two. With so few trials per group, and the authors calling all subgroup work exploratory, that is a lead to follow up, not a recommendation to skip the bike.

Host 1And the cholesterol results?

Host 2Eight trials and two hundred and fifty-three women each for HDL and LDL. HDL, the cholesterol usually linked to lower cardiovascular risk, rose with a g of zero point eight four, interval zero point two four to one point four five. LDL, the one linked to higher risk, fell with a g of minus zero point seven eight, interval minus one point two nine to minus zero point two seven.

Host 1Total cholesterol and triglycerides?

Host 2Nine trials each, and neither was statistically significant. Total cholesterol came in at minus zero point three five, with an interval just touching zero, and triglycerides at minus zero point two four, with an interval that also allows a small effect in the wrong direction. Those two were downgraded twice for imprecision.

Host 1Anything odd in the lipid subgroups?

Host 2The HDL and LDL improvements looked much larger in trials of women aged forty and over, and in women with comorbidities like metabolic syndrome or high blood pressure. In the four trials of women with polycystic ovary syndrome, the HDL effect was close to zero, at zero point one three. But these comparisons involve two to five trials each, so they cannot separate age from disease from trial quality.

Who seemed to benefit

Host 1Did they find anything about who benefits most?

Host 2One finding goes against intuition. For body mass index, trials where women started below thirty showed a g of minus zero point nine, while trials starting at thirty or above showed minus zero point one seven, which was not significant. That subgroup difference was statistically significant, and a meta-regression found the same: the higher the average starting index, the smaller the effect.

Host 1So the heavier the women, the less it helped?

Host 2That is what the numbers say at the level of trials, and the same pattern showed up for fat mass. But it compares trials, not women, so it may reflect who those trials recruited, how long they ran, or how well they were run. A possibility I would add myself: a standardised effect shrinks when the spread between women is wider, which it may be in heavier groups, so the same kilograms could look smaller.

Host 1What about how often people trained?

Host 2Trials with more sessions per week showed larger fat reductions. For fat mass, training frequency appeared to explain all of the variation between trials. But only two trials trained more than three times a week, the simple subgroup comparison was not significant, and the authors themselves warn that the explained variance is unstable.

Why it matters

Host 1Pull it together. What is the principle here?

Host 2That relatively short programmes of hard intervals, usually three sessions a week for a few months, reliably make women with overweight or obesity fitter than doing nothing, and probably trim some body fat. The fitness gain is the part the evidence supports most consistently.

Host 1Who should care?

Host 2Clinicians and trainers who work with women with obesity, including women with polycystic ovary syndrome, after menopause or after breast cancer treatment, since all of those groups appear in these trials. It is reasonable evidence that interval training is one usable option for them, and that its main payoff is fitness.

Host 1And what does it not mean?

Host 2It does not show that intervals beat walking, jogging or any other exercise, because no trial compared them. It does not show that intervals alone will produce large weight loss. And it says nothing about blood pressure, blood sugar or long-term heart disease, which were outside the review.

Host 1Anything about safety?

Host 2Nothing that can be pooled. Adverse events, attendance and how much training women actually completed were reported too inconsistently, and the authors explicitly say that missing reports of harm are not evidence that there was no harm.

Honest caveats

Host 1Now the caveats properly.

Host 2First, the authors' own grading: low certainty for body weight and very low for every other outcome. In the GRADE system, very low means the true effect is likely to be substantially different from the estimate. That comes from the authors, not me.

Host 1Second?

Host 2The end-of-training comparison. In tiny trials, two groups can start out quite different, and comparing only final values carries that gap into the result. Looking at the body weight forest plot, one trial ended with the interval group at about seventy-five kilograms and the controls at about eighty-four, and another at sixty-seven versus seventy-six, which is a lot for eight to twelve weeks of exercise.

Host 1So those gaps were there from the start?

Host 2Probably in part, but that is my reading, not the paper's. Gaps can also run the other way: in one trial the interval group ended heavier than the controls. With fifteen trials, those imbalances may partly cancel out, but it makes each pooled number less precise than the confidence interval suggests.

Host 1Third?

Host 2Size and quality. Interval groups had eight to thirty women each, the largest analysis covers three hundred and seventy women, no trial was at low risk of bias, and four were at high risk. Small trials of this kind tend to overestimate effects.

Host 1And publication bias?

Host 2They found no clear sign of it, but with eight to twelve trials per outcome the tests have little power, and the authors say a negative test does not show bias is absent. Add the databases they did not search and the five reports they could not retrieve.

Host 1Anything else?

Host 2The mix of populations: college students, women with polycystic ovary syndrome, cancer survivors, women with high blood pressure, and women in their sixties. The pooled number averages over all of them. And our copy of the paper ends mid-results, so I have not seen the authors' sensitivity checks for fitness and cholesterol, nor any statement on funding.

What to watch next

Host 1So what should we watch for next?

Host 2Larger trials in women with obesity that randomise to intervals, to steady moderate exercise and to no exercise, so the question of which exercise is answered rather than whether to exercise. Preregistered, with change from baseline as the main outcome, and with attendance and adverse events reported properly.

Host 1And beyond twelve weeks?

Host 2Longer follow-up, to see whether women keep doing intervals after a supervised programme ends, since adherence is the argument for intervals in the first place. And a direct test of whether heavier women really respond less, using individual participant data rather than trial averages.

Sign-off

Host 1That is the deep dive: Effects of high-intensity interval training on body composition, cardiorespiratory fitness, and selected lipid outcomes in adult women with overweight or obesity, by Yulong Wang, Ming Li and colleagues, in Frontiers in Physiology. The link is in the show notes.

Host 2Short version: across fifteen small trials, interval training versus no exercise left women clearly fitter, a little lighter and somewhat less fat, with better HDL and LDL cholesterol. The certainty is low to very low, and nothing here says intervals beat other kinds of exercise.

Host 1Every Friday this same feed carries the roundup episode: five new papers as trailers, and you pick which become deep dives. This has been the Hastrup Review of Science.