Measurement of biological age with Levine’s Phenotypic Age calculator is strongly correlated with chronological age (r=0.94; see https://michaellustgarten.wordpress.com/2019/09/09/quantifying-biological-age/). In 2019, I measured all 9 of its analytes 3 times, with biological age readings of 35.39y, 35.58y, and 31.3y, for an average 2019 biological age of 34.09y (see https://michaellustgarten.wordpress.com/2019/11/01/biological-age-31-3y-chronological-age-46y/). That’s 12 years younger than my chronological age in 2019, 46y!
On Feb 12, I had my first blood test measurement of 2020. I expected to see a worse biological age, as over the past week, I’d been hit with the flu, and since my last measurement in 2019, I made a few changes to my diet that I didn’t expect to favorably affect it. In contrast, I’ve been purposefully in a mild caloric restriction in an attempt to reduce my body fat from a relatively lean 10-12% to lower values. Since my last blood test 3+ months ago, my average calorie intake was 2553, which is 5-10% less than my maintenance intake, 2700-2800 calories/day. So how did these variables affect my biologic age? Let’s have a look at the data!
My biological age was 32.75y, which is less than my 2019 average value, and better than I expected considering the factors mentioned above! Note that there is room for improvement, including my creatinine and WBC levels, which both increased when compared with my average 2015-2019 values (which included 23 blood tests). My average daily fiber intake has been ~100g/day for a few years, and over the past 3 months, I purposefully reduced that to ~70g/day. Conversely, I increased my intake of meat, eggs, and cheese intake during that period, to see if eating less fiber and more animal products would negatively impact my blood test results. For me, eating more animal protein and less total fiber may not be optimal, as my creatinine levels also rose in 2019 when I performed a similar dietary experiment. Note that creatinine levels increase with age (see https://michaellustgarten.wordpress.com/2019/11/18/optimizing-biologic-age-creatinine/), so if I can avoid that by altering my diet, I will. For the next blood test, I’ll reduce, but not eliminate my intake of meat, eggs, and cheese, and I expect that my creatinine levels will decrease back towards my average 2015-2019 value of 0.94 mg/dL.
Also note my WBCs-although they’re not higher than the 3.5-6 optimal range (see https://michaellustgarten.wordpress.com/2019/10/11/blood-testing-whats-optimal-for-wbc-levels/), they’re increased when compared with my average 2015-2019 value of 4.5. This increase is more than likely a result of the flu/infection that I’m battling. Once it passes, I expect it to return to close to my average WBC value, ~4.5.
Going forward, I expect my creatinine and WBCs to come down to their average values, which would result in a biological age that is closer to 30y on my next blood test. Stay tuned for that data!
To quantify your biological age using Levine’s Phenotypic Age calculator, here’s the Excel link! DNAmPhenoAge_gen (1)
If you’re interested, please have a look at my book!
Thank you. Interesting. I would expect you will keep those 12 for the coming years but still on rising slope as the chronological age is still factored in Levine’s. Would be interesting to compare with aging.ai and other methodologies non including age. I bet aging.ai gives you more than 12 y younger.
Ha, I can see why you would think that, but I have room for improvement. With another 6-months to ~a year of mild CR, I expect further improvements, bringing me closer to 20 years younger than my chronological age. Also note that based on this measurement, I’m 14y younger than my chronological not 12, which was the 2019 average reduction.
Yep, aging.ai is indeed younger.
Thanks for a great discussion. Indeed all these biological clocks seem lightly correlated. Per epigenetic clocks. There were not enough older people in the previous samples, so cpg saturation makes clocks errors greater in those cohorts. “The age prediction properties of both Horvath  and Hannum et al.  DNA methylation clock models begin to degrade as subjects enter old age. This is at least partly due to saturation, i.e., DNA methylation proportion at some loci approaching 0 or 1, and confounding with the effects of other age-related processes will also play a role. It is likely that this could be ameliorated with additional loci and/or further refined modeling of the currently used set.”https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1810-4
“…centenarians are younger (8.6 years) than expected based on their chronological age.” https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4712339/
Also Michael said that the Levine clock was cheaper to do. I am using another company to test my epigenome, https://trumelabs.com/ for $99 so this will be the 3rd test and is fairly inexpensive.
Yes, you can get epigenetic clock testing done for a decent price. But it isn’t DNAmGrimAge, which is the best of the bunch. Note that it’s likely the earlier iterations of DNAm testing, which didn’t identify correlations between smoking with epigenetic age, which is ridiculous.
This is such a lovely blog with some fantastic articles and I really get a lot from it so please keep up the great work. If you are at all interested this link will take you to a summit that i attended in realtion to health and wellbeing in the later years and I found it so insightful. Take a look here: http://article.media/aging1
There are many biological clocks and disparity of ages which reduces their utility. Why don’t you compare and contrast a few epigenetic clocks, VO2Max, Levines Phenotypic age. I found a 38 year difference between my VO2Max and epigenetic ages, greater than 50% of my chronological age.
Yes, there are many biological clocks, but the analytes on Levine’s PA calculator have been studied for 50-100y+. In contrast, the data on epigenetic clocks is growing, but they’ve been studied for ~10 years, and they’re not all the same. The initial iteration was not correlated with smoking, which is ridiculous. The latest, and best of the bunch is DNAmGrimAge, but that’s not commercially available.
In terms of VO2 max, I haven’t done that in a while, but in case you missed it, I’ve been slowly reducing my resting heart rate, while increasing my HRV, which would suggest a higher VO2 max:
And my muscle strength is pretty good, too:
Actually the epigenetic clock has better correlation to age vs Levine clock at r=.96 vs .94 . There are issues with both these types of clocks such as sample size for each phenotype that result in inaccuracies. Elite athletes in Poland epigenetic age is 5-6 years older than chronological age which may be an artifact of small sample size for example. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5842850/ Also there are various types of aging, at least 7, and these clocks use the immune cells which may not be the complete story w/r/t all forms of aging.
“The epigenetic clock leads to a chronological age prediction that has a Pearson correlation coefficient of r=0.96 with chronological age ” https://en.wikipedia.org/wiki/Epigenetic_clock
Levine says ” In this population, phenotypic age is correlated with chronological age at r=0.94″ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5940111/
I get that 0.96 is > 0.94, but is it that 0.02 difference much better? That’s debatable. Plus Levine’s PA test costs much less, and as I mentioned, it’s analytes have been studied for a long time. I know how each of those analytes changes with age, and with mortality risk, in case you missed it:
Levine’s Phenotypic age test covers biomarkers representative of many organ sytems and cell types, including the liver (albumin), kidney (creatinine), WBCs, RBCs, inflammation.
Also, I hadn’t seen that paper, good stuff, thanks for that.
Thanks Michael. GrimAge seems the best biological clock (per “DNAm GrimAge stands out among existing epigenetic clocks in terms of its predictive ability for time-to-death (Cox regression P=2.0E-75), time-to-coronary heart disease (Cox P=6.2E-24), time-to-cancer (P= 1.3E-12” – https://www.aging-us.com/article/101684/text ) and I don’t think there is yet a commercial version of it. Epigenetic clocks don’t say why there is an age discrepancy because deep learning doesn’t explain itself. I’ve used mydnage.com x 2 and trueme.com x 1, but they seem to be using older Horvath clocks. There is a great discrepancy between the age results of all these biological clocks that I don’t fully understand except to say they are based on different data sets and small samples. I think it would be interesting if you compared and contrasted all these clocks, to reveal the confidence we should place in them despite our hopes for the future.
VO2Max on Garmin fitness watches is computed by First Beat, “Firstbeat can automatically detect your VO2max fitness level during walking and running activities, using a proprietary method shown to be 95% accurate compared to laboratory measurements. ” https://www.firstbeat.com/en/consumer-feature/vo2max-fitness-level/“, but I can not find a correlation to age. Studies show how various factors like HIIT and mitochondrial senescence affect it.
As you mentioned, the commercially available DNAm clocks are using older versions, so I’ll wait for DNAmGrimAge to come out.
In terms of VO2 max, not only is the absolute # important, but also making sure it doesn’t decline during aging. I’d rather have a VO2 max of 45 knowing I can maintain that until I’m 100, rather than 60 and declining every year.
I stand with Stuart in particular with ” ..There is a great discrepancy between the age results of all these biological clocks that I don’t fully understand except to say they are based on different data sets and small samples…” I recollect also a paper by Belsky et al (https://doi.org/10.1093/aje/kwx346) saying just that. I also stand with Michael as those clinical biomarkers are used since long in the clinic. I would compare at least Levine’s , aging.ai and one methodology not using chronological age (e.g. Mitnitski’s or one version of Klemera-Doubal)
The correlation for aging.ai with biological age (r=0.80) isn’t as good as Levine’s PA. Nonetheless, independent of these aging-related clocks, knowing how these analytes (that have been studied for a long time) change with age and mortality risk goes beyond simple BA calculation. That’s 1 reason that I’ve written about many of them in lit reviews on my site…
Pingback: 15+ Years Younger Than My Chronological Age: Blood Test #2 In 2020 – Michael Lustgarten