Analyze biological feedback

What they measured vs what we can measure

The single most useful table on this site. For every endpoint used in the original work, it says where it was measured (cells, mice, humans, or only asserted in a patent) and whether a DIY self-experimenter can measure the same thing, with what.

Legend: yes routinely doable · partly proxy or specialist lab · no needs tissue, autopsy or a research facility. Sources: Aging Cell 2025 [S1], mouse data [S1], NCT07168525 [S8], NCT06562374 [S7], patent application US20240001155A1 [S3].

The markers, in three tiers

Tier 1 · Orderable today in an Italian private lab, no prescription needed

Not available in Italian routine labs, so they belong to Tier 2: Activin A, osteopontin, TNFR1, MMP-2/7/9, PAI-1, CCL3, IL-15, RAGE, CCL18, p16 in T cells, flow SA-β-gal, γH2AX, quantitative cfDNA, suPAR.

Tier 2 · With a partner research lab

• p16^INK4a + p21 + cGAS RT-qPCR on sorted CD3+ T cells — the closest thing to a "senescent burden" blood test (Liu & Sharpless 2009: ~10× rise over six decades; exercise lowers it). 10–15 mL EDTA, CD3 microbeads, RNA, TaqMan. Consumables roughly €40–90 per sample. Any university molecular-biology lab can do it.
• Flow-cytometry SA-β-gal on PBMCs (CellEvent Senescence Green, ~$350 for 50 tests, or SPiDER-βGal) with surface markers: the senescent-like fraction is highest in CD8+ cells. Needs fresh PBMCs within hours.
• SASP multiplex: Luminex/MSD custom panel or Olink Target 96 Inflammation (92 proteins from 1 µL). Decisive advantage for n-of-1: all pre and post samples run on the same plate, which removes inter-assay variance. Olink's practical minimum is a full plate, so it only makes sense with several participants.
• RNA-seq on PBMC/CD3+ scored with SenMayo (125-gene senescence set) — the home-brew analogue of the "RNA-seq" in the UT protocol.
• EPIC v2 methylation array at an EU service, clocks computed with open software (DunedinPACE, pyaging). This is the only serious way to use clocks in an n-of-1, because you can run technical duplicates and put pre/post samples on the same chip.

Tier 3 · Mail-in, consumer

Urine, honestly

There is no validated urinary SASP panel. 8-oxo-dG and 8-oxo-Gsn measure whole-body oxidative repair turnover, not senescence, with strong diurnal variation and ELISA-vs-mass-spectrometry disagreement. F2-isoprostanes are a solid oxidative-stress marker with no direct link to senescent burden. Collect urine only if a partner lab has LC-MS/MS; otherwise it does not deserve budget. It is the weakest link in the protocol.

Why repetition matters more than the marker

For a marker with analytical variation CV_A and within-person biological variation CV_I, a change only counts if it exceeds the reference change value: RCV ≈ 2.77 × √(CV_A² + CV_I²). Averaging k samples before and k after shrinks it by √(1/k) — with k = 3 the requirement drops by 42%.

Read this against the expected effect size. Interventions with real senolytic or lifestyle effects move single analytes by 10–30% in studies with dozens of subjects. So in an n-of-1: GDF-15 is the only credible primary endpoint; IL-6 and hs-CRP alone cannot support a conclusion; everything else is exploratory. Report every marker as "Δ% with its RCV beside it". "−18% with RCV ±12.5%" is a result. "IL-6 −25% with RCV ±62%" is not.

Timing: two traps specific to ultrasound

Rule: chronic-effect samples are drawn at least 48–72 hours after the last session, always at the same interval. Optionally add an explicitly exploratory acute series (pre-session, +2 h, +24 h) on one single session.

And remember the bath itself. Forty-five minutes in 35 °C water is an intervention: haemodynamics, IL-6, sleep and mood all respond. The UT trial has a sham arm for precisely this reason. An n-of-1 has no sham, so the washout phase is the closest substitute, and alternating "bath without ultrasound" sessions is better still.

The n-of-1 protocol

1. Standardise everything before you measure anything. Same time of day (07:30–09:00), 12 h fasting, no intense exercise for 48 h (GDF-15 rises 4–5× after a marathon), no infection, vaccination or surgery in the previous 4 weeks, no alcohol for 48 h, same tube type and same lab every time, centrifuged within 1–3 h, aliquoted and frozen at −80 °C.
2. Three baselines, not one (weeks −4, −2, 0), each with the full functional battery repeated three times. This is also how you estimate your own biological variation.
3. Eight weeks of exposure, matching the UT trial: 45 min, 3×/week. Log every session with its measured acoustic dose.
4. Three post samples (weeks 8, 9, 10), the first 72 h after the last session.
5. Washout re-test at week 16. A return toward baseline is the strongest evidence obtainable without a sham arm.
6. Freeze spare aliquots — your personal biobank costs nothing and lets you re-run everything in one batch later, which is what removes inter-assay noise.
7. Pre-declare one primary endpoint (proposed: plasma GDF-15, mean of 3 pre vs mean of 3 post) and label everything else exploratory. Publish raw data and code so that many identical n-of-1 runs can later be pooled.

Functional battery (aligned with NCT07168525)

Every functional test: 3 repetitions per session, 3 baseline sessions. That gives your own measurement error, and the detectable change is 1.96 × √2 × that error.

Panels and costs

Kits and reagents worth naming

Oncological safety monitoring

Stop rule: any persistent abnormality — cytopenia, monoclonal peak, rising LDH, weight loss, a persistent lymph node — means stop and see a doctor, not "re-measure next month".

Full working notes with sources, Italian price lists, variability data and unverified items: 04 — biomarkers and Italian lab availability (in Italian).