Drift Detection: Catching Change Before It Becomes a Problem
The changes that matter most are too gradual to notice. Detecting them requires a kept record and knowing which four signals move earliest.
The Short Answer
Almost nothing consequential in health announces itself. Blood pressure rises by a few millimetres a year, apoB drifts upward within the reference range, fasting insulin creeps, grip strength declines, and training volume falls without any decision to reduce it. Each of those is invisible month to month and substantial over a decade. Detecting drift requires two things: a record that permits comparison, and knowing which signals move earliest.
Why Drift Is Hard to Notice
The rate is below perceptual threshold. A change of one or two per cent per year is not detectable by how you feel, and by the time it is, a decade of it has accumulated.
Adaptation. Reduced capacity is accommodated by doing less, which removes the signal. Someone whose aerobic capacity has declined avoids stairs and no longer notices being out of breath.
Reference ranges are wide. A marker can move substantially within its range without ever being flagged, which is why a normal result is not the same as an unchanged one.
Memory is unreliable for gradual change. People consistently misremember their own baseline, in both directions.
Attribution to age. Gradual decline gets attributed to getting older, which is partly true and prevents asking whether it is modifiable. A substantial share of apparent age-related decline is deconditioning.
No single event. There is no moment at which the change happened, so there is nothing to prompt attention.
The combination means the changes with the largest cumulative consequence are precisely the ones least likely to be noticed without a record.
The Four Earliest Signals
| Signal | What it detects | Why it is early |
|---|---|---|
| Heart rate at a fixed submaximal workload | Aerobic capacity | Sensitive, cheap, and precedes subjective change |
| Waist circumference | Visceral adiposity | Moves before weight in some cases; tracks the metabolically active compartment |
| Fasting insulin and triglyceride to HDL ratio | Insulin resistance | Years before fasting glucose or HbA1c move |
| Sleep regularity | Circadian and behavioural drift | Changes first, and drives several downstream markers |
Heart rate at a fixed workload is the most useful single item on this list, because it costs nothing beyond consistency, it responds within weeks, and it precedes the subjective sense that fitness has declined. The same pace at a higher heart rate than a year ago is a real signal.
The insulin row is the classic case of a late marker masking an early process. Compensated hyperinsulinaemia can persist for years with normal glucose, which means a person reassured by HbA1c is being reassured about a later stage.
Sleep regularity belongs here because it is upstream: it drifts first, and it degrades glycaemic control, inflammatory markers, blood pressure and recovery downstream.
Reading Within-Range Movement
The most useful interpretive habit in this whole area is reading position within a range rather than whether a value was flagged.
GGT within its normal range associates with cardiovascular events, diabetes incidence and mortality in large cohorts. A GGT drifting from the low end to the upper end is a real signal and never gets flagged.
Apolipoprotein B exposure is cumulative over decades, so a value drifting upward within range is accumulating risk even while reported as normal. Reference ranges describe a population in which cardiovascular disease is the leading cause of death.
Creatinine drifting upward within range over five years is more informative than a single flagged value, particularly with cystatin C for confirmation.
Red cell distribution width rises early in iron shortfall, often before MCV falls or ferritin is flagged.
Fasting insulin has reference ranges extending well into insulin-resistant territory, so within-range is not reassurance.
The general rule: record the number, not the flag. A sequence of numbers permits trend detection; a sequence of "normal" does not.
What Makes a Record Usable
A record only permits comparison if the conditions were comparable, which is where most personal records fail.
Same laboratory. Between-assay differences can exceed real biological change, particularly for insulin, free testosterone and thyroid antibodies.
Same conditions. Fasting state, time of day, recent exercise, recent illness. A panel drawn three days after a hard session is not comparable to one drawn rested.
Same time of year. Several markers vary seasonally, vitamin D most dramatically.
Context recorded alongside. Illness, training, travel, sleep, medications and cycle day. Without them a deviation is uninterpretable and invites an unnecessary response.
Raw numbers with units. A score or a colour cannot be compared with a clinical result or read by a clinician.
In one place. A sequence spread across three apps and a drawer of paper reports is not a sequence.
Long enough. Two points establish a difference; five or more establish a trend and reveal the size of ordinary variation, which is what makes a genuine change recognisable.
Distinguishing Drift From Noise
This is the practical difficulty, since every marker varies for reasons unrelated to any underlying change.
Require direction over multiple points. Three consecutive annual values moving the same way is a trend; two differing values is a difference.
Check whether related markers agree. Rising fasting insulin with rising triglycerides, rising waist and rising GGT is coherent. An isolated change in one marker with everything else stable is usually context or measurement.
Check the context first. Illness, hard training, travel, poor sleep, weight change and a new medication explain most single deviations.
Know the measurement variability. Triglycerides vary substantially between draws; apoB much less. A given percentage change means different things for each.
Repeat before acting. Particularly for anything that would trigger a change.
Do not respond to a single point. The most common error, and the one that produces protocol churn.
Do respond to a consistent direction, even when every value is within range. That is precisely the case a flag-based system misses and a record catches.
What Detecting Drift Is Worth
The value is entirely in the interval between detection and consequence, and that interval can be decades.
Apolipoprotein B exposure is cumulative, so detecting an upward drift at 40 rather than at 60 changes twenty years of exposure. Blood pressure identified in midlife matters more for cognitive outcomes than blood pressure identified at 75. Insulin resistance caught in the compensated phase is more reversible than after glucose has risen. Declining grip strength and gait speed reverse with training, and the earlier the intervention the higher the starting point.
Against that, the cost of over-detection is real: reacting to noise produces unnecessary investigation, protocol churn and anxiety. The discipline of requiring a consistent direction across multiple points is what separates useful drift detection from hypervigilance.
The practical conclusion is unglamorous and specific. Measure a small number of reliable things, under consistent conditions, at sensible intervals, record the numbers rather than the flags, keep them in one place, and read direction over years. That practice costs almost nothing and it is the only way the changes that matter most become visible while they are still cheap to address.
The AEONNN Perspective
Drift detection is the practical form of AEONNN's premise that biology is never still. The changes with the largest cumulative consequence are below perceptual threshold, accommodated by doing less, and hidden inside wide reference ranges, which means they are invisible without a kept record.
The platform reads four signals as earliest: heart rate at a fixed submaximal workload, waist circumference, fasting insulin with the triglyceride to HDL ratio, and sleep regularity. The first costs nothing beyond consistency and precedes subjective change. The third is years ahead of glucose and HbA1c, which is why a member reassured by HbA1c is being reassured about a later stage.
The interpretive habit that matters most is reading position within a range rather than whether a value was flagged, since GGT, apoB, creatinine and fasting insulin all carry information inside their reference ranges. That is only possible if the number is recorded rather than the flag, under comparable conditions, in one place, over enough points to distinguish direction from ordinary variation. The Quality layer supplies those conditions, and without them a member's own sequence loses most of its value.
Pillar Matrix mapping
Database Matrix layers
- Real-Time User Layer (wearable and adherence signals)
- Population Layer (UK Biobank, NHANES)
- Quality / Formulation Layer (ConsumerLab, Labdoor)
- Evidence Layer (PubMed, Cochrane, ClinicalTrials.gov)
Frequently Asked
Why is gradual change so hard to notice?
The rate is below perceptual threshold, reduced capacity gets accommodated by doing less, reference ranges are wide, memory is unreliable for gradual change, and there is no single event to prompt attention.
Which signals drift earliest?
Heart rate at a fixed submaximal workload, waist circumference, fasting insulin with the triglyceride to HDL ratio, and sleep regularity. All four precede subjective change.
Why is HbA1c a late signal?
Because compensated hyperinsulinaemia can persist for years with normal glucose. Fasting insulin captures the compensation itself, which is why it moves earlier.
Does a normal result mean nothing changed?
No. A marker can move substantially within its range without being flagged, and GGT, apoB, creatinine and fasting insulin all carry information inside their reference ranges.
What makes a personal record usable?
Same laboratory, same conditions, same time of year, context recorded alongside, raw numbers with units, kept in one place, and enough points to distinguish direction from variation.
How do I tell drift from noise?
Require a consistent direction across three or more points, check whether related markers agree, check the context first, know the marker’s variability, and repeat before acting.
What is early detection worth?
The interval between detection and consequence, which can be decades. ApoB exposure is cumulative, midlife blood pressure matters more for cognition than late-life, and compensated insulin resistance is more reversible.
Evidence and review
Any dosage ranges cited here reflect the ranges used in published human trials, not personal recommendations. Evidence in this field moves, so this article is reviewed quarterly and carries its last-updated date above. Nothing here is intended as medical advice, and supplementation should be discussed with a qualified clinician, particularly alongside prescribed medication or an existing condition.