Kidney Clocks, Stopping Sets Early and an Eccentric Black Hole

Sponsor

0:03Host 1Support for the Hastrup Review of Science comes from RefurbMiner.

0:07Host 1One intelligent platform for your whole business. Invoicing, banking, and compliance in a single system, with every decision fully traceable. Privacy ready, and hosted in Europe.

0:19Host 2Start your free fourteen day trial today. The link is in the show notes.

Cold open

0:24Host 1Welcome to the Hastrup Review of Science — where artificial intelligence reads real science, so you don't have to pretend you understood the paper.

0:34Host 1This week's lead: a German cohort sampled the same four thousand people up to three times, and their protein-based biological age moved with their kidneys, their weight and their cholesterol.

0:46Host 2If an aging clock tracks change within one person, it starts to look like something you can measure an intervention against, not just a label.

Proteomic aging clock

0:55Host 1Start there. What did they build?

0:58Host 2The KORA cohort in Augsburg ran mass spectrometry plasma proteomics on four thousand one hundred people at up to three visits, about seven years apart. That gives nine thousand one hundred and five samples, ages twenty-five to eighty-eight. A neural network predicts age with a mean error of six point four years, and the gap between predicted and actual age is the age acceleration.

1:22Host 1And the within-person result?

1:24Host 2When a person's own kidney function rose one standard deviation, their age acceleration fell by about one point eight years. Rising body mass index added about one point three years, and crossing into high LDL added about point nine. Kidney function was estimated from markers that were not inputs to the clock, so the link is not circular.

1:45Host 1Does it predict anything hard?

1:47Host 2Mortality, over up to twenty-two years of follow-up, but the gain is small: the concordance index went from point eight two five with age alone to point eight three zero. The caveats are clear. It is a preprint, it is lower accuracy than the biggest clocks, there is no independent replication cohort, and within-person models still miss confounders that change over time.

2:10Host 1The full story is in this paper's deep dive, out this week in the feed.

Nutri-Dense meals trial

2:15Host 1Next, a diet trial from Japan that also measured epigenetic age.

2:20Host 2Three hundred adults were randomised for four weeks to one of three meal plans. One was optimised to the Japanese reference intakes for twenty-six micronutrients, one was a micronutrient-insufficient control, and one was a low-carbohydrate, high-fat version with the same micronutrients.

2:38Host 1What happened on the main outcomes?

2:41Host 2Nothing. Visceral fat area and salivary cortisol and melatonin did not differ between groups. On secondary outcomes, LDL cholesterol fell in both optimised arms against control, and systolic blood pressure fell in the low-carbohydrate arm.

2:57Host 1And the biological age result?

2:59Host 2That came from a subgroup of older men, ten per group, where the optimised arm showed lower PhenoAge. Ten per group over four weeks is a hypothesis, not a finding, and the authors say so. We had only the abstract for this one, so we are not adding detail beyond it.

3:16Host 1The deep dive is out this week in the feed.

How much protein

3:20Host 1Now to training, starting with protein.

3:23Host 2Stuart Phillips of McMaster replies to two critics of his line that one point six grams per kilo per day is a ceiling for lifters. He discloses up front that he co-authored much of the evidence under dispute, and that his own position has moved from a range of one point six to two point two.

3:41Host 1What is the core argument?

3:43Host 2Trials measure lean mass, not muscle. In twenty-one women scanned the same morning, scanner lean mass exceeded muscle measured by creatine dilution by about twenty kilos, roughly half. Taken at face value, three months of training adds one point one kilos of fat-free mass, and supplemental protein adds point three.

4:04Host 1And going above one point six?

4:06Host 2On the most generous reading, about two hundred grams of lean mass from one point six to two point two, and head-to-head trials against intakes up to three point two found no clear difference. The honest caveat runs both ways: it is a non-peer-reviewed commentary, and the small, short trials that limit his critics limit him too.

4:27Host 1The full story is in this paper's deep dive, out this week in the feed.

Sponsor break

4:32Host 2Do you have problems with a dry, itchy beard? Then you are in luck.

4:37Host 2Beard oil conditions the hair and the skin beneath it, so the beard sits soft and smooth from the chin down. It is in stock right now, in a wide range, and very reasonably priced.

4:48Host 1I use it every morning. A few drops, worked in slowly, and the whole thing just relaxes.

4:53Host 2Unscented options are available for those who prefer discretion. Warm a little between the palms and apply with long, firm strokes.

5:01Host 1It also makes the beard much easier to handle, which my partner has remarked on.

5:06Host 2Our referral link is in the show notes.

Velocity loss and sprint

5:09Host 1Velocity loss, and what it means for sprinters.

5:13Host 2Velocity loss is how much bar speed you let a set drop before you stop. This preregistered three-level meta-analysis pooled twelve randomised trials and forty-two effect sizes in trained people, comparing lower against higher thresholds.

5:27Host 1The result?

5:29Host 2Lower thresholds won by a Hedges' g of point two nine for jump and sprint combined, with essentially no heterogeneity and a prediction interval that stayed above zero. The countermovement jump effect was point two nine, and ten percent loss against thirty percent gave point four three. Twenty against forty percent was small and not significant.

5:50Host 1How much should we trust it?

5:52Host 2Certainty was low to moderate, and every included trial had some concerns for risk of bias. Samples were small, most programmes ran four to eight weeks, and the participants were mostly trained men, so masters athletes and women are an extrapolation. The practical reading is modest: stopping sets earlier costs nothing for speed and may help a little.

6:14Host 1The full story is in this paper's deep dive, out this week in the feed.

GW231123 eccentricity

6:20Host 1To physics. The heaviest black hole merger on record, looked at again.

6:24Host 2GW231123 has component masses near one hundred and thirty-seven and one hundred and one solar masses, inside or at the edge of the pair-instability gap where single stars should not leave black holes. This team re-fit it with an eccentric, precessing waveform and used Hanford data down to eleven hertz instead of the usual twenty.

6:45Host 1What did they find?

6:46Host 2An eccentricity of about point three four, and a Bayes factor of ten point two for the eccentric model over the circular one. With eccentricity included, the second black hole's spin drops from about point nine to about point seven, which is what you expect from the remnant of an earlier merger.

7:05Host 1Why did nobody see this before?

7:07Host 2Their injection tests show that starting the analysis at twenty hertz washes the eccentricity out. The caveats: a Bayes factor of ten is meaningful but not decisive, it rests on one waveform family and on glitch-subtracted data from one detector, and it is an unreviewed preprint.

7:24Host 1The full story is in this paper's deep dive, out this week in the feed.

Virgo infall and H0

7:29Host 1Last, the Hubble constant measured from galaxies falling into Virgo.

7:34Host 2Near a massive cluster, galaxy motions are not noise: they are infall you can model. The team tested the standard infall formulas on simulated halos in IllustrisTNG, where the true answer is known, and found the common fitting functions bias the Hubble constant low.

7:51Host 1And applied to Virgo?

7:52Host 2With one hundred and twenty-one tip-of-the-red-giant-branch distances, the preferred model gives seventy-four point four, plus or minus two point nine, statistical only. Correcting for the bias measured in simulations raises it to eighty-one point six, but the error grows to plus or minus seven point eight.

8:12Host 1So does it settle the tension?

8:14Host 2No. The direction sides with the high local values, but the corrected uncertainty is too wide to discriminate, and the correction depends on how well the simulation matches the real Virgo. Note also that Adam Riess, a leading voice for the high value, is a co-author.

8:30Host 1The full story is in this paper's deep dive, out this week in the feed.

Sign-off

8:36Host 1That is the Hastrup Review of Science for this week.

8:39Host 2Read the caveats, then read the papers.