Deep divefrom Episode 5Longevity5 October 202623 min

Five Years Younger in Six Months? Reading the Blood Clock

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: Alenezi, Sabico et al. — Structured lifestyle modification and biological aging: effects on PhenoAge, sclerostin, and GDF-15 in Arab adults with prediabetes (Frontiers in Aging)

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. One paper today: can six months of diet, walking and counseling make a person with prediabetes biologically younger, and what is actually inside that number?

Host 2The paper is a randomized trial by Ahmed Alenezi, Shaun Sabico and colleagues at King Saud University in Riyadh and North Medical Tower Hospital in Arar, Saudi Arabia, with Nasser Al-Daghri as the senior author. It was published on September tenth, twenty twenty-six, in Frontiers in Aging, after peer review. The headline is that the lifestyle group's biological age estimate dropped by almost five years in six months, while the control group's rose by just over one.

Host 1Five years younger in six months is the kind of claim that should make everyone a little suspicious.

Host 2It should, and by the end of this episode we will have taken that number apart piece by piece. The trial itself is real and the lifestyle program clearly worked on weight and blood sugar. What the biological age change means is a more interesting question than the headline suggests.

Why biological age exists

Host 1Start with the idea. Why would anyone want a biological age that is different from the age on your passport?

Host 2Because two sixty-year-olds can have very different bodies. One has the blood pressure, kidneys and blood sugar of a fifty-year-old, the other is already frail, and chronological age alone cannot tell them apart. As established background, researchers have been trying to compress that difference into a single number for decades.

Host 1How do you even build that number?

Host 2There are two broad families. Epigenetic clocks, like the one Steve Horvath published in twenty thirteen, read chemical marks on DNA called methylation, which shift in predictable ways as we age. Clinical clocks instead combine ordinary blood test results that any hospital lab already produces.

Host 1And PhenoAge is one of the clinical ones.

Host 2Yes. It was published in twenty eighteen by Morgan Levine and colleagues, who started from forty-two candidate biomarkers in a large American health survey called NHANES three and asked which combination best predicted death over the following years. A penalized regression narrowed it to nine blood markers plus chronological age.

Host 1Which nine?

Host 2Albumin, creatinine, glucose, C-reactive protein, the lymphocyte percentage, mean red cell volume, red cell distribution width, alkaline phosphatase and the white blood cell count. Each gets a weight, the weighted sum becomes a predicted mortality risk, and that risk is converted back into the age at which an average person would carry it.

Host 1So PhenoAge is really a mortality risk score wearing an age costume.

Host 2That is a fair description, and it matters later. A PhenoAge five years younger means your blood panel looks like that of someone with a lower predicted mortality risk. It does not by itself mean any cell in your body has aged differently.

What was known before

Host 1Before this trial, what did we know about lifestyle and diabetes risk?

Host 2As background, the lifestyle side is very well established. The American Diabetes Prevention Program, published in two thousand two, found that an intensive lifestyle program cut new cases of diabetes by about fifty-eight percent compared with placebo in people with prediabetes, and the Finnish Diabetes Prevention Study found a similar reduction. The program in this paper is built on that template.

Host 1And what about lifestyle and biological age?

Host 2Much thinner, and the authors say so. Most evidence is observational, people who exercise and eat well tend to have younger biological age scores, which cannot separate cause from who those people already are. The paper cites two trials that used PhenoAge: the CALERIE caloric restriction trial, which saw a promising but non-significant slowing, and the DO-HEALTH trial, which reported protective effects of omega three, vitamin D and exercise.

Host 1Both in healthy people.

Host 2Right, and that is the authors' central argument. If healthy volunteers have little room to improve, the effect might be larger in people already on a path toward metabolic disease, such as adults with prediabetes. They also point out that the Middle East is underrepresented in this research, despite high rates of obesity and diabetes there.

Host 1The title also names two molecules. What are they?

Host 2Growth differentiation factor fifteen, which I will call GDF fifteen, is a stress-signaling protein whose blood levels rise with age, inflammation and mitochondrial trouble, and track with mortality. Sclerostin is made mostly by osteocytes, the cells buried inside bone, and it acts as a brake on bone building by blocking a signaling pathway called Wnt.

Host 1Why would a diabetes trial measure a bone protein?

Host 2As background, sclerostin has turned up in links between bone, fat, muscle and blood sugar, and a drug that blocks it, romosozumab, is approved for osteoporosis. The authors treated it as a window onto the skeletal and metabolic side of aging, while openly saying they did not know which direction it should move.

The open question

Host 1So what is the precise question?

Host 2Can a structured six-month lifestyle program, of the type already used in Saudi primary care, change PhenoAge in Arab adults with prediabetes, compared with a control group. The primary endpoint was PhenoAge, decided in advance, and GDF fifteen and sclerostin were secondary endpoints. The trial was registered at ClinicalTrials dot gov before results were known.

Methods: who and how

Host 1Walk me through what they did. Who was in it?

Host 2Saudi adults aged thirty or older, recruited through primary healthcare centers in the Arar region in the north of the country. Prediabetes was defined by the American Diabetes Association cutoff, a hemoglobin A one C between five point seven and six point four percent, which is a roughly three-month average of blood sugar. People with diabetes, pregnant or breastfeeding women and non-Saudis were excluded.

Host 1How many people?

Host 2Four hundred and twelve were screened, two hundred and thirty-one did not meet the criteria, and one hundred and eighty-one were randomized. Ninety-three went to the control group and eighty-eight to the lifestyle program.

Host 1What did the typical participant look like?

Host 2Average age about fifty-one, and a body mass index of thirty, which is the threshold for obesity. One hundred and thirty were men and fifty-one were women, so about seventy-two percent male. Average A one C was five point nine percent, and among the participants with medical history data, half had hypertension and almost three quarters had a family history of type two diabetes.

Host 1And their PhenoAge going in?

Host 2Forty-nine point one years on average, slightly younger than their actual average age of fifty point eight. So this was not a group that started out looking biologically old on this measure.

Methods: the program

Host 1What did the lifestyle group actually have to do?

Host 2The targets were taken from earlier Saudi diabetes prevention studies. Lose at least five percent of body weight, do at least one hundred and fifty minutes of moderate activity per week, keep fat under thirty percent of calories and saturated fat under ten percent, and eat at least fifteen grams of fiber per thousand calories.

Host 1And the support?

Host 2Individual dietary counseling with a registered nutritionist every two weeks, six structured education sessions twice a month over the first three months, and online support throughout. That is a substantial amount of contact over six months.

Host 1And the control group?

Host 2Printed educational material and an appointment to come back in six months. So the comparison is an intensive, personal program against a brochure, which is a fair test of the program as a whole, but it cannot tell you which ingredient mattered.

Host 1How was randomization done?

Host 2With a sequence generated by the random function in Microsoft Excel. The authors state plainly that there was no additional allocation concealment, meaning the people enrolling participants could in principle know the next assignment. The paper does not mention blinding of staff or outcome assessors either.

Methods: measurements

Host 1What did they measure?

Host 2At baseline and at six months: weight, height, waist, hip and neck circumference, and blood pressure. Fasting blood went to the Arar Central Hospital lab for the routine tests, glucose, A one C, albumin, creatinine, alkaline phosphatase, a full blood count and C-reactive protein. Frozen serum was shipped to King Saud University in Riyadh for GDF fifteen and sclerostin, measured with commercial assay kits.

Host 1And PhenoAge itself?

Host 2Calculated from those routine results with a publicly available spreadsheet implementation of Levine's formula, which they downloaded on October twenty-sixth, twenty twenty-five and cite for reproducibility. That is a nice touch: anyone can check the arithmetic.

Host 1And the statistics?

Host 2Repeated measures analysis of variance, which asks whether the two groups changed differently over time, called the time by group interaction. Skewed variables were log-transformed, and no covariates were adjusted because the groups looked balanced at baseline. Every randomized participant was analyzed, and the paper reports that nobody was lost to follow-up in either arm.

Host 1Nobody dropped out in six months?

Host 2Zero out of one hundred and eighty-one, according to the flowchart. That is unusual for a lifestyle trial, and the results section phrases it as one hundred and eighty-one participants who successfully completed the intervention. The paper does not explain how complete retention was achieved.

Results: PhenoAge

Host 1Now the main result.

Host 2The lifestyle group's PhenoAge went from forty-eight point four to forty-three point five years, an average drop of four point nine years, with a ninety-five percent confidence interval from minus six point two to minus three point seven. The control group went from forty-nine point eight to fifty-one point one, an increase of one point three years. The interaction was highly significant, p below zero point zero zero one, with a reported effect size of zero point two one seven.

Host 1So the gap between groups is about six years.

Host 2Six point two years, measured over six months of calendar time. The spread was wide too: the standard deviation of the change was six point three years in the lifestyle group and five point six in controls, so individual results varied enormously.

Host 1The control group aged one point three years in half a year?

Host 2On this score, yes, and the confidence interval of zero point one to two point five excludes zero, with a within-group p of zero point zero three five. Oddly, the text describes that rise as not statistically different from baseline, which contradicts its own numbers. We will come back to why the control group's score climbed.

A word from our sponsor

Host 2Do you ever look in the mirror and notice the passing years are most visible in your nostrils? Then you are in luck.

Host 2The Hastrup Review of Science is brought to you by nose-hair trimmers. A small rotating blade, a rounded guard, and a quiet motor that does one job and does it well.

Host 1I use mine every morning. It takes about five years off, although the effect size varies a lot between individuals.

Host 2They come battery or rechargeable, many with washable heads, and several fit comfortably in a travel bag. Widely available and very reasonably priced.

Host 1And it gets right up in there without anybody needing to know.

Host 2Nose-hair trimmers. Our referral link is in the show notes.

Results: weight, sugar, blood pressure

Host 1Back to the trial. Did the program change the ordinary health measures?

Host 2Clearly. The lifestyle group's average weight fell from eighty point two to seventy-four point five kilograms, a loss of five point seven kilograms, while the control group gained one point six. Body mass index fell by two points, waist by six point three centimeters, and A one C from five point nine to five point five percent, while the controls drifted up to six point zero.

Host 1And blood pressure?

Host 2Systolic pressure fell about four and a half points in the lifestyle group and rose about two and a half in controls, and diastolic followed the same pattern. All of these interactions had p values at or below zero point zero zero three.

Host 1One thing in the abstract: it calls the five point seven kilograms a between-group difference.

Host 2Those abstract figures, five point seven kilograms, two body mass index points, six point three centimeters of waist and zero point four percent A one C, are the lifestyle group's own change from baseline. The difference against controls is actually larger for weight, around seven kilograms, because the controls gained. It is a labeling slip, not a hidden problem, but it is worth knowing when you read the summary.

Results: adherence

Host 1Did people actually follow the program?

Host 2Partly. Seventy-eight point four percent of the lifestyle group reached the five percent weight loss target, against four point three percent of controls. Twenty-one point six percent hit the fat target versus nine point seven percent, and fiber goal achievement was similar in both groups, around a third.

Host 1And the exercise target?

Host 2Median walking time went from zero to sixty minutes in the lifestyle group, while the controls stayed at zero, but the authors note the group still fell short of the one hundred and fifty minute weekly goal. Median energy intake in the lifestyle group fell from about one thousand nine hundred and eighty to about one thousand seven hundred and ninety calories a day.

Host 1So most of the weight loss came from eating less rather than moving more.

Host 2That is the most plausible reading of these numbers, though the paper did not measure activity in detail, and it says so as a limitation.

Results: GDF fifteen and sclerostin

Host 1Now the two molecules. What happened to GDF fifteen?

Host 2It fell in both groups. The control median dropped from seven hundred and fifty to about five hundred and ninety picograms per milliliter, and the lifestyle median from about seven hundred and ninety to about five hundred and sixty-four. The lifestyle drop was somewhat larger, but the interaction p value was zero point six, so there is no evidence the program caused it.

Host 1And sclerostin?

Host 2It rose in both groups. The control median went up by about one hundred picograms per milliliter, from one hundred and thirty-six to two hundred and thirty-seven, while the lifestyle median rose by only about thirteen. That difference looks large, but the interaction p value was zero point eight, because the spread within each group was enormous.

Host 1So both secondary endpoints are null.

Host 2Yes, and the authors report them honestly as time effects rather than intervention effects. With only two measurements and large variability, a fall in GDF fifteen in both arms could reflect season, sample handling or assay batch as much as biology, and the authors themselves note that two time points cannot separate intervention responses from ordinary fluctuation.

Host 1Was there any link between these molecules and PhenoAge?

Host 2At baseline, GDF fifteen correlated fairly strongly with PhenoAge, with an r of zero point five six, which fits its reputation as an aging and stress marker. Sclerostin had a weak inverse correlation. But changes in either molecule over six months did not track changes in PhenoAge.

Host 1Anything odd in those correlations?

Host 2One small thing. For sclerostin they report a change correlation of minus zero point three five seven with a p value of zero point zero seven, and with one hundred and eighty-one people a correlation that size would be far more significant than that. It suggests the analysis ran on a smaller subset, which the paper does not describe.

What moved the clock

Host 1Here is the question I care about. What actually moved the PhenoAge needle?

Host 2Albumin in the lifestyle group rose from 3.8 to 4.9 g/dL, while in the control group it fell from 4.2 to 3.8 g/dL.

Host 1Why does albumin matter so much?

Host 2In Levine's formula, higher albumin means lower predicted mortality, and it carries a heavy weight. As our own back-of-the-envelope check using the published PhenoAge weights, not something the paper reports, a rise of one point one grams per deciliter in albumin alone is worth roughly four years of PhenoAge. Across the two groups, albumin accounts for roughly half of the six-year gap, the glucose improvement for roughly a quarter, and a rise in creatinine among controls, which was not itself significant, for much of the rest.

Host 1So weight loss itself is not in the formula.

Host 2Correct, PhenoAge never sees weight, waist or blood pressure directly. It only sees blood markers, so the question becomes whether albumin really rose by more than a gram per deciliter in people who lost weight.

Host 1Is that plausible?

Host 2As background, albumin is a fairly stable protein, and a typical reference range tops out around five grams per deciliter. A group average of four point nine with a standard deviation of one point six implies a meaningful share of follow-up values well above anything usually seen in healthy adults. The paper does not discuss this, so a lab or calibration change between visits, or a data issue, cannot be ruled out, and it is the first thing I would ask the authors about.

Host 1Anything else strange in the blood tests?

Host 2C-reactive protein, the inflammation marker, roughly tripled in both groups, from a median of two point five to around seven and a half milligrams per liter. The baseline median sits exactly at the lower quartile, which hints at a reporting floor, and a simultaneous tripling in both arms looks more like a measurement shift than a biological change. Because PhenoAge also uses C-reactive protein, these shifts feed straight into the score.

Host 1So the control group's one point three year rise might partly be the lab, not aging.

Host 2That is possible, and the same would apply to part of the lifestyle group's drop. The randomized comparison protects against shifts that hit both groups equally, but albumin moved much more in one group than the other, and that is exactly where the effect comes from.

Why it matters

Host 1Step back. Why does this matter?

Host 2The principle the trial supports is modest and sensible: if you improve someone's metabolic health, a composite risk score built from their blood tests improves too. That is consistent with decades of diabetes prevention research and adds a randomized data point from a population that is rarely studied.

Host 1Who should change what?

Host 2For someone with prediabetes, the practical message is the old one, now replicated in Arar. A structured program with regular nutritionist contact got nearly four in five participants to five percent weight loss and brought average A one C back to five point five percent, which is the part of this trial I would trust most.

Host 1And what does it not mean?

Host 2It does not mean anyone became five years younger. PhenoAge was built to predict mortality across a population, and there is, as background, no evidence yet that pushing an individual's score down by changing albumin produces the matching drop in mortality. The authors themselves are careful here, describing the result as a change in a composite clinical measure and not in epigenetic aging.

Host 1And the molecules?

Host 2For GDF fifteen and sclerostin this trial offers nothing actionable. They moved over time in both groups, and the program did not detectably change either.

Honest caveats

Host 1Let us collect the caveats in one place.

Host 2First, design. Randomization by spreadsheet without allocation concealment, no reported blinding, and an attention-heavy program compared with a brochure, all of which can inflate benefits on softer outcomes. The blood tests themselves are less vulnerable to expectation, which helps.

Host 1Second?

Host 2Population. Saudi adults from one northern region, about seventy-two percent men, with a baseline PhenoAge already a little younger than their real age, which the authors acknowledge limits generalization. Medical history was available for only ninety participants and family history for one hundred and four.

Host 1Third?

Host 2Duration and endpoints. Six months, two measurements, and no hard outcomes such as how many people developed diabetes. A sustained benefit would need longer follow-up, since weight regain after lifestyle programs is common.

Host 1And the measurement questions.

Host 2Fourth, and most important for the headline: the PhenoAge drop leans heavily on an albumin rise that is unusually large and unusually variable, alongside a tripling of C-reactive protein in both arms, and none of this is discussed. There are also small reporting inconsistencies, the abstract labeling, the control group's significant rise described as not different, and the sclerostin correlation. None of these proves anything is wrong, but together they say the six-year figure should be held loosely.

Host 1Any conflicts of interest?

Host 2None declared. The work was funded by a research chair program at King Saud University, and the authors state that no generative artificial intelligence was used in writing it.

What to watch next

Host 1What would you watch for next?

Host 2A release of individual-level data, or at least the distribution of albumin values at each visit and whether the lab changed methods between them. Recomputing PhenoAge without albumin, or with it held at baseline, would show how much of the effect survives.

Host 1And beyond this trial?

Host 2Trials that measure a clinical clock and an epigenetic clock side by side, with more than two time points and longer follow-up, and that report actual diabetes onset. The authors plan to add telomere length in future studies, and they call for repeated measurements of GDF fifteen and sclerostin to see how they really behave.

Sign-off

Host 1That is the deep dive: Structured lifestyle modification and biological aging, by Alenezi, Sabico and colleagues, in Frontiers in Aging. The link is in the show notes.

Host 2Every 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.