eduKate Learning Manual
Science | Veterinary World
Establish the Individual Baseline → Separate Analytical Noise From Biological Change → Compare the New Result With the Previous Result → Check Context → Look for Direction → Reassess the Animal

Veterinary Biological Variation and Serial Testing

Why a Result Can Stay “Normal” While the Animal Is Changing

Wait, What? Two Results Can Both Sit Inside the Reference Interval and Still Tell a Very Different Story

Imagine a cat whose creatinine concentration has been steady for years near the lower part of a laboratory reference interval. At the next health check, the value is still technically “normal”, but it has risen substantially from that cat’s previous pattern. Nothing has crossed the population boundary. Yet the animal may have changed.

This is where serial testing becomes more interesting than a red flag beside a single number. A reference interval compares an animal with a population. A trend compares the animal with itself.

normal for the population ≠ unchanged for the individual.

The Scientific Job

This manual owns one Veterinary World job:

How do veterinary teams decide whether a change between serial laboratory measurements is more likely to reflect real change in the animal rather than ordinary biological fluctuation or measurement variation?

Veterinary Reference Intervals owns population-based comparison. Diagnostic Tests owns general test performance. Cardiac Biomarkers, Urinalysis, Endocrine Testing and other manuals own their analytes and disease contexts. This page owns change across time within the same animal.

Quick Answer

A serial result is interpreted against more than one source of variation:

• Analytical variation: the measurement system itself is not perfectly identical every time.
• Within-animal biological variation: healthy physiology moves around an individual homeostatic set point.
• Between-animal variation: different healthy animals can have different typical values.
• Preanalytical variation: sampling, timing, handling and patient state can alter the specimen.
• True clinical change: disease, treatment, ageing, hydration, nutrition or another biological process may shift the animal’s state.

The question is not simply whether the latest value is high or low. It is whether the magnitude and direction of change are larger than expected from ordinary variation and whether the animal’s clinical story supports that interpretation.

Primary Entry — Every Animal Has a Moving Baseline

Healthy physiology is dynamic. Hormones pulse. Hydration changes. Activity varies. Feeding alters metabolism. Immune activity rises and falls. Laboratory measurements therefore fluctuate even when an animal is not becoming ill.

That does not make laboratory testing unreliable. It means reliability includes knowing how much movement is ordinary.

Part 1 — Population Reference Intervals Answer a Different Question

A population reference interval asks where most values from a defined healthy reference population fall under a specified method. It is useful for identifying results that are unusual compared with that population.

But an individual animal may naturally occupy only a narrow part of that interval. If so, a large personal change can occur before the result crosses the population boundary.

eduKate Veterinary World — Veterinary Reference Intervals

Part 2 — Within-Animal Variation Is Not Random Chaos

Many analytes fluctuate around a reasonably stable individual centre over a defined period. Researchers can estimate this within-subject variation by repeatedly sampling clinically stable animals under controlled conditions.

The size of this variation differs between analytes. A change that is striking for one measurement may be unremarkable for another.

Part 3 — Analytical Variation Adds Another Layer of Movement

No measurement process is perfectly identical from run to run. Even a well-controlled analyser produces small differences because reagents, calibration, temperature and instrument performance vary within accepted limits.

Serial interpretation therefore has to distinguish a genuine biological shift from the combined effect of ordinary biological fluctuation and analytical imprecision.

Part 4 — Reference Change Values Estimate How Big a Difference Needs to Be

A reference change value, sometimes described through related terms such as critical difference, combines estimates of analytical and within-subject biological variation. It provides a statistical way to ask whether two serial measurements differ more than would commonly be expected from those sources of variation alone.

This is not a diagnosis threshold. It is a change threshold. Crossing it suggests that a real shift becomes more plausible, not that one specific disease has been proven.

Secondary Deepening — Direction Often Matters More Than One Isolated Number

A single unexpected value may be noise, transient physiology or early disease. Three measurements moving steadily in the same direction create a different evidential shape. Trends can reveal acceleration, plateau, response or relapse.

That is why longitudinal medicine is not simply “repeat the test”. It is interpret the pattern of repeated tests.

Part 5 — Serial Monitoring Works Best When Conditions Are Comparable

A morning fasting sample and an evening post-treatment sample may differ for reasons unrelated to disease progression. Different laboratories, instruments or assay methods can also create shifts.

Comparable sampling conditions make longitudinal interpretation stronger because fewer alternative explanations remain.

Part 6 — Individuality Explains Why Some Population Intervals Are Less Helpful for Monitoring

Research in cats and dogs shows that some analytes display substantial individuality: healthy animals differ from one another more than an individual animal fluctuates around its own typical level. In those situations, serial change can carry information that a broad population interval misses.

Other analytes vary more within the same animal, making population intervals comparatively more useful. The balance is measurement-specific.

Part 7 — “Stable” Does Not Mean Identical

A stable animal can produce different numbers on different days. Stability is a range of expected variation, not a frozen value.

This prevents overreaction to tiny changes. When every movement is treated as disease, serial monitoring creates noise instead of insight.

Part 8 — “Normal” Does Not Mean Reassuring in Every Context

If a value changes substantially from the animal’s previous pattern, the fact that it remains inside a population interval does not erase the trend. Conversely, a mildly out-of-range value that has remained stable for years in a healthy individual may have a different meaning from a sudden new abnormality.

The laboratory number gains meaning from history.

JC Deepening — Serial Testing Is a Problem of Signal Detection

We can think of a longitudinal measurement as a signal moving through noise. The signal is the clinically meaningful change we care about. The noise includes biological fluctuation, preanalytical effects and analytical imprecision.

useful trend = observed change interpreted against expected variation.

Reducing noise improves sensitivity to change. That is why consistent sampling, reliable methods and good records matter.

Part 9 — Different Analytes Need Different Expectations

Studies of canine NT-proBNP, cardiac troponin, urinary analytes, serum chemistry and cortisol demonstrate that biological variability is not uniform. Some markers require relatively large changes before serial differences become persuasive. Others are more tightly regulated.

There is therefore no universal percentage change that means “clinically significant” for every laboratory measurement.

Part 10 — Longitudinal Data Can Reveal the Animal Before Disease Becomes Obvious

Routine health checks may create a valuable personal history. When future illness develops, previous results provide context for what “usual” looked like for that individual. The value of a baseline is often realised only later.

This is one reason careful record continuity matters in veterinary medicine: yesterday’s ordinary result can become tomorrow’s comparator.

How Do We Know?

Veterinary clinical pathology studies estimate analytical variation, within-subject variation, between-subject variation, indices of individuality and reference change values by repeated sampling. Recent work has extended these methods to large clinical databases as well as controlled studies. The evidence consistently shows that meaningful serial interpretation depends on the specific analyte and the magnitude of expected variation.

Observation vs Inference

• Observation: a dog’s NT-proBNP has risen between two visits.
• Inference: cardiac change may be more plausible, but biological and analytical variation must be considered.
• Observation: a cat’s creatinine remains inside the reference interval but has increased from its long-term baseline.
• Inference: a within-individual shift may deserve attention; kidney disease is not established by trend alone.
• Observation: a third measurement continues in the same direction.
• Inference: a persistent biological process becomes more plausible than a single isolated fluctuation.

Evidence Boundaries

• inside the reference interval ≠ unchanged for the individual.
• outside the reference interval ≠ disease automatically present.
• large change ≠ one specific diagnosis.
• small change ≠ no disease possible.
• serial trend ≠ proof of causation.
• reference change value ≠ treatment threshold.
• repeat testing ≠ identical biological conditions.
• educational longitudinal reasoning ≠ an individual monitoring schedule.

Common Misconceptions

Misconception	Better model
If the result is normal, nothing changed.	The animal can move substantially within a broad population interval.
Any change between tests is clinically meaningful.	Biological and analytical variation create expected movement.
One percentage threshold works for every analyte.	Variation differs substantially among measurements.
More tests always create more certainty.	Poorly standardised serial tests can create more noise.

Unfamiliar Transfer

Animal A has one mildly high result but no previous data. Animal B has three “normal” results that rise steadily over a year. Animal C has two very different measurements taken at different laboratories under different conditions. Animal D has a large change that returns to baseline on repeat testing.

A strong learner asks not only “is it normal?” but “how much did it change, how much change is expected, and were the measurements truly comparable?”

Checkpoint Questions

1. What is the difference between a population reference interval and an individual trend?
2. What is within-subject biological variation?
3. Why does analytical variation matter in serial testing?
4. What question does a reference change value help answer?
5. Why can a normal result still be concerning?
6. Why should serial samples be collected under comparable conditions?
7. Why is one universal percentage-change rule unsafe?
8. How can routine historical results become valuable later?

Answer key

1. A population interval compares with other healthy animals; an individual trend compares the animal with itself.
2. It is the ordinary physiological fluctuation around an individual’s typical level.
3. The analyser itself contributes expected measurement variation.
4. Whether the difference between two results is larger than expected from biological and analytical variation.
5. It may have shifted substantially from that animal’s baseline while remaining inside the population interval.
6. Comparable conditions reduce alternative explanations for change.
7. Different analytes have different biological and analytical variability.
8. They establish the individual’s previous pattern before disease appears.

Edge Science — Can Longitudinal Models Learn the Animal’s Personal Normal?

As veterinary records accumulate, statistical and computational systems may estimate an individual animal’s expected range rather than relying only on population limits. Such systems could flag unusual personal shifts earlier.

The danger is overconfidence. A personal model must account for method changes, ageing, missing data, illness episodes and the fact that a baseline itself can drift. The best future system will treat longitudinal prediction as evidence, not verdict.

Veterinary World Direction Graph

Veterinary serial testing → individual baseline → comparable sample conditions → measurement → expected biological variation + analytical variation → magnitude of change → direction across time → clinical context → specific specialist owner → reassessment.

Reference Intervals owns population comparison. Specific diagnostic manuals own their measurements. This page owns interpretation of change across time.

Research Sources and Further Reading

• Biological variation of baseline cortisol in healthy dogs — clinical applications
• Biological variation of biochemical analytes estimated from a large database of dogs and cats
• Biological variation of urine and serum analytes in healthy dogs
• Biologic variability of NT-proBNP in healthy dogs and dogs with mitral valve disease
• Haematologic and biochemical biologic variation in cats
• eduKate Veterinary World — Veterinary Reference Intervals

Educational safety boundary: The significance of serial changes depends on the analyte, laboratory method, clinical context and the animal’s history. This manual does not define an individual patient’s diagnostic or monitoring thresholds.

Teaching Guide for Parents, Tutors and Teachers

For the people who teach because somebody depends on them.

Draw a wide “normal population” box and then draw a much narrower personal line moving inside it. Ask the learner whether the personal line can shift meaningfully without leaving the box. That visual usually reveals the whole idea.

compare with the population → compare with the individual → estimate expected variation → watch direction → return to the animal.

The mastery target is a learner who stops reading laboratory reports as isolated red and black numbers and begins to see measurements as a time series connected to a living animal.