The Laboratory Tests Medicine Knows About but Rarely Uses Well

Modern medicine has an odd problem.

It is simultaneously capable of astonishing diagnostic sophistication and surprisingly repetitive routine testing.

A patient can have a genome sequenced, a tumor profiled molecule by molecule, an antibody measured against a single protein, or a liver’s mechanical stiffness estimated without a biopsy.

And yet a typical outpatient visit may still revolve around roughly the same familiar package:

• CBC
• CMP
• hemoglobin A1C
• TSH
• conventional lipid panel
• perhaps a urinalysis

There is nothing wrong with those tests. They are extraordinarily useful.

The problem is that they can create the illusion that if those numbers look reassuring, we have looked everywhere worth looking.

We have not.

There is a category of laboratory medicine that sits in an unusual middle ground. These tests are not experimental. They are not the proprietary mega-panels marketed by wellness companies. They are not based on vague ideas about “toxins,” “adrenal fatigue,” or individualized supplement optimization.

They are recognized by mainstream professional societies, grounded in established physiology, and supported by substantial evidence.

But they are often absent from routine workflows, ordered only after years of disease, or interpreted less intelligently than they could be.

This article is about that gap.

It is also about something larger than the tests themselves.

The more interesting failure is sometimes not which analyte we measure, but how we think about measurement.

A laboratory report encourages a deceptively simple model:

$$\text{measurement} \rightarrow \text{reference interval} \rightarrow \text{normal or abnormal}$$

Clinical physiology is rarely that simple.

A better model is:

$$\text{clinical meaning} = \text{measurement} + \text{baseline} + \text{trajectory} + \text{related measurements} + \text{pretest probability} + \text{measurement limitations} + \text{clinical context}$$

That difference—between checking numbers and testing physiologic hypotheses—is the central idea of this article.

Important note: The case studies below are fictional composites. Their laboratory values and clinical situations are intentionally realistic, but they do not describe specific patients and are not instructions for self-diagnosis or treatment.

---

First, what does “rarely uses well” mean?

It would be easy to turn this subject into an argument that physicians should simply order more laboratory tests.

That would be a mistake.

Every test has costs:

• financial cost
• false positives
• incidental findings
• follow-up procedures
• anxiety
• biological variability
• analytical limitations
• the possibility of treating a number rather than a patient

A test is valuable when it answers a meaningful question and can reasonably change what happens next.

So when we call a test “underused,” we should mean something more specific:

1. The test measures a physiologically important phenomenon.
2. There is reasonable evidence connecting that measurement to clinically meaningful risk, diagnosis, prognosis, or treatment.
3. Major guidelines or expert bodies recognize its use in an identifiable population.
4. The information cannot always be recovered from the ordinary routine panel.
5. Knowing the result can plausibly alter a decision.

That definition eliminates a great deal of fashionable laboratory testing.

What remains is much more interesting.

---

1. Lipoprotein(a): the cardiovascular risk factor hiding outside the lipid panel

Most adults eventually become familiar with a conventional lipid panel:

• total cholesterol
• LDL cholesterol
• HDL cholesterol
• triglycerides

But cardiovascular risk is not completely described by those four numbers.

One important omission is lipoprotein(a), usually written Lp(a).

Lp(a) is an LDL-like particle with an additional apolipoprotein(a) attached to apolipoprotein B. Its concentration is strongly genetically determined and usually remains relatively stable through adulthood.

That makes it unusual.

Diet, exercise, and weight loss can dramatically improve many cardiovascular risk factors. They generally have much less effect on Lp(a).

The 2026 ACC/AHA dyslipidemia guideline recommends measuring Lp(a) at least once in adulthood. The guideline identifies values around 125 nmol/L or 50 mg/dL and above as a risk-enhancing range and notes substantially greater risk at very high concentrations.

That is a striking recommendation when compared with ordinary practice: the test is not part of the standard lipid panel, yet for most adults it only needs to be measured once to expose a major inherited piece of cardiovascular biology.

Case study: the reassuring LDL

Patient: Elena, 48

Elena runs several times per week, does not smoke, has normal blood pressure, and has no diabetes.

Her lipid panel looks unremarkable:

Test	Result
Total cholesterol	196 mg/dL
LDL-C	116 mg/dL
HDL-C	61 mg/dL
Triglycerides	94 mg/dL

Nothing is spectacularly abnormal.

But her father had a myocardial infarction at age 49, and an uncle had coronary bypass surgery in his early fifties.

Her clinician orders Lp(a).

Lp(a): 238 nmol/L

The interpretation changes immediately.

The LDL result was real. It simply did not describe all of Elena’s atherogenic risk.

The Lp(a) result does not diagnose coronary artery disease and does not mean that Elena is destined to have a heart attack. What it does is reveal an inherited risk enhancer that makes aggressive control of the modifiable risk factors more valuable.

Depending on the complete risk assessment, that may strengthen the rationale for lower LDL exposure, meticulous blood-pressure control, smoking avoidance, diabetes prevention, and—when the decision remains uncertain—additional risk stratification such as coronary artery calcium imaging.

The important lesson is not “Lp(a) is better than LDL.”

It is:

A normal-looking conventional panel can be incomplete because it measures only part of the causal system.

What Lp(a) is good for

• identifying inherited cardiovascular risk not visible in an ordinary lipid panel
• explaining some families with premature atherosclerotic cardiovascular disease
• refining preventive decisions
• prompting cascade consideration in relatives when markedly elevated

What Lp(a) is not good for

• explaining every cardiovascular event
• replacing LDL-C
• serving as a general marker of lifestyle quality
• being repeatedly measured without a specific reason when the value is already known and stable

Evidence anchor: The 2026 ACC/AHA guideline recommends at least one Lp(a) measurement in adulthood and describes Lp(a) as a risk-enhancing factor. See the American College of Cardiology summary of the 2026 dyslipidemia guideline.

---

2. ApoB: counting atherogenic particles instead of only measuring their cholesterol cargo

The conventional LDL-C value asks roughly:

How much cholesterol is being carried inside LDL-class particles?

But two people can carry the same amount of cholesterol inside very different numbers of particles.

That distinction matters because apolipoprotein B, or ApoB, is present as one structural molecule on essentially each major atherogenic lipoprotein particle.

ApoB therefore acts as a practical estimate of the number of atherogenic particles, not merely the cholesterol mass they contain.

This becomes particularly useful when cholesterol content and particle number are discordant.

That can happen in people with:

• hypertriglyceridemia
• type 2 diabetes
• metabolic syndrome
• cardiovascular-kidney-metabolic disease
• established cardiovascular disease despite apparently acceptable LDL-C

Case study: two LDLs that are not physiologically equivalent

Consider two fictional patients with the same LDL-C:

LDL-C: 92 mg/dL

Patient A has:

• triglycerides 82 mg/dL
• ApoB 76 mg/dL

Patient B has:

• triglycerides 246 mg/dL
• ApoB 118 mg/dL

The LDL cholesterol concentration is identical.

The particle biology is not.

Patient B is carrying the cholesterol in a larger number of ApoB-containing particles. That discordance can reveal residual atherogenic risk that LDL-C alone understates.

This is an important general principle:

Concentration and count are not the same biological variable.

The 2026 ACC/AHA dyslipidemia guideline supports selective ApoB measurement to refine risk in several groups, including people with diabetes, high triglycerides, cardiovascular-kidney-metabolic syndrome, or known cardiovascular disease.

ApoB does not need to replace the lipid panel to be valuable. It solves a specific problem the standard panel sometimes cannot solve well.

Evidence anchor: See the American College of Cardiology’s 2026 dyslipidemia guideline summary.

---

3. Aldosterone and renin: hypertension as a hormonal phenotype

One of the most interesting underused laboratory paradigms is not a single test at all.

It is a relationship between two hormones.

Aldosterone promotes sodium retention and potassium excretion. Renin participates in the upstream regulatory system that normally stimulates aldosterone production when the body needs it.

Under ordinary physiology, the two exist in a feedback relationship.

In primary aldosteronism, aldosterone production becomes partially autonomous.

So a characteristic biochemical pattern appears:

$$\text{aldosterone inappropriately high} + \text{renin suppressed}$$

The ratio between them—the aldosterone-to-renin ratio, or ARR—can therefore expose a hormonal cause of hypertension that an ordinary CMP cannot.

This matters because primary aldosteronism is not merely “high blood pressure with a fancy diagnosis.” It can carry greater cardiovascular and kidney risk than ordinary primary hypertension, and it can be treated specifically with mineralocorticoid-receptor blockade or, in appropriately selected unilateral disease, surgery.

In 2025, the Endocrine Society substantially broadened its guideline and suggested screening all individuals with hypertension for primary aldosteronism, subject to implementation constraints.

That recommendation itself tells us something about the historical implementation gap.

Case study: waiting for low potassium

Patient: Marcus, 45

Marcus has hypertension despite two medications.

His potassium is:

K = 3.8 mmol/L

That is within the laboratory reference interval.

Historically, a clinician might think:

Primary aldosteronism causes hypokalemia. Potassium is normal. Move on.

But hypokalemia is not required.

Marcus undergoes appropriately prepared aldosterone-renin screening.

• renin: markedly suppressed
• aldosterone: inappropriately elevated relative to that renin
• ARR: clearly elevated for the laboratory’s assay

The result is repeated under controlled conditions because medications, sodium intake, potassium status, time of day, posture, kidney disease, and assay methodology can all affect interpretation.

Further evaluation ultimately supports primary aldosteronism.

Marcus is not simply given a third nonspecific antihypertensive and sent home. His treatment is now directed at the mineralocorticoid pathway responsible for the physiology.

The important laboratory lesson is subtle:

The abnormality may live in the relationship between two values even when neither value, viewed alone, appears spectacular.

Why the ARR is easy to misuse

The ARR is not a magic number.

A very low renin value can mathematically inflate a ratio. Aldosterone assays differ. Antihypertensive medications can raise or suppress renin. Hypokalemia can suppress aldosterone and generate a false-negative screen.

The Endocrine Society explicitly emphasizes local assay cutoffs, interfering medications, potassium, pretest probability, and repeat testing when appropriate.

So this is simultaneously an argument for more screening and for more sophisticated interpretation.

Evidence anchor: The 2025 Endocrine Society Clinical Practice Guideline on Primary Aldosteronism suggests screening all people with hypertension using aldosterone and renin, with potassium measured alongside them to aid interpretation.

---

4. UACR: kidney disease can exist while creatinine still looks fine

Routine chemistry panels have trained generations of patients to look at one kidney number:

creatinine

Modern reports usually add an estimated glomerular filtration rate, or eGFR.

Those are important.

But kidney disease has more than one dimension.

A useful simplification is:

• eGFR asks how well the kidneys are filtering.
• albuminuria asks whether the filtration barrier is leaking albumin.

A person can have important glomerular injury while filtration remains relatively preserved.

The urine albumin-to-creatinine ratio, or UACR, is therefore one of the most consequential measurements that can be absent from a routine “kidney check.”

Case study: the normal creatinine that was not enough

Patient: Denise, 57

Denise has type 2 diabetes and hypertension.

Her annual CMP shows:

• creatinine: 0.82 mg/dL
• eGFR: 84 mL/min/1.73 m²
• potassium: normal

She is relieved.

“My kidneys are normal.”

But a urine albumin-to-creatinine ratio is also collected.

UACR: 386 mg/g

A repeat measurement confirms persistent, markedly increased albuminuria.

Now the picture is very different.

Her kidneys are still filtering reasonably well, but the glomerular barrier is leaking a substantial amount of albumin. That changes kidney-risk classification and cardiovascular-risk assessment and can affect the urgency and selection of kidney-protective therapies.

If we had looked only at creatinine, we would have asked only one kidney question.

The deeper principle

Many organs fail along multiple axes.

For the kidney:

$$\text{kidney status} \neq \text{filtration alone}$$

A better approximation is:

$$\text{kidney risk} \approx \text{filtration} + \text{albumin leakage} + \text{trajectory} + \text{etiology} + \text{clinical context}$$

KDIGO’s modern CKD framework explicitly incorporates both GFR category and albuminuria category.

Evidence anchor: The current global CKD standard remains the KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease, which emphasizes updated measurement of both estimated GFR and albuminuria.

---

5. Cystatin C: when creatinine is telling us as much about muscle as kidney function

Creatinine is useful partly because the body produces it continuously and the kidneys filter it.

But creatinine production depends substantially on skeletal muscle.

That means a creatinine-based eGFR is not purely a kidney measurement.

It is a kidney estimate filtered through the patient’s creatinine generation.

This becomes especially important in people with unusual body composition:

• frailty
• very low muscle mass
• amputation
• paralysis
• severe chronic illness
• some forms of liver disease
• unusually high muscularity

Cystatin C provides an alternative filtration marker with a different set of non-GFR determinants. When greater accuracy is needed, equations combining creatinine and cystatin C can outperform creatinine alone.

Case study: excellent eGFR in a patient with very little muscle

Patient: Harold, 76

Harold has lost substantial muscle after months of illness.

His creatinine is:

0.68 mg/dL

The creatinine-based eGFR appears excellent.

But the result is physiologically suspicious. Harold has little muscle from which to generate creatinine.

Cystatin C is measured.

The cystatin-C-based estimate is much lower, and the combined creatinine-cystatin C equation places kidney function in a range that materially changes how a renally cleared medication is dosed.

The lesson is not that cystatin C is intrinsically “better” than creatinine.

It is that two biomarkers with different biases can be combined when one biomarker’s assumptions are clearly violated.

That is a powerful general diagnostic strategy.

When a measurement may be distorted by a known confounder, seek an independent measurement with a different error structure.

Evidence anchor: KDIGO’s 2024 CKD guideline incorporates modern use of creatinine, cystatin C, and combined equations when more accurate GFR assessment is needed. See KDIGO CKD Evaluation and Management.

---

6. FIB-4: using ordinary laboratory values to detect extraordinary liver risk

Some underused tests require ordering an unfamiliar assay.

FIB-4 does not.

The Fibrosis-4 index uses values many patients already have:

• age
• AST
• ALT
• platelet count

A commonly used form is:

$$\mathrm{FIB\text{-}4} = \frac{\text{age} \times \text{AST}} {\text{platelets} \times \sqrt{\text{ALT}}}$$

Its purpose is not to diagnose every liver disease.

It is primarily a risk-stratification tool for advanced fibrosis.

That distinction matters because aminotransferases can be deceptively reassuring.

A patient can accumulate important liver fibrosis without an ALT of 300 or an AST of 500.

Case study: “My liver enzymes are basically normal”

Patient: Robert, 55

Robert has type 2 diabetes, obesity, hypertension, and hypertriglyceridemia.

His laboratory results include:

• AST: 34 U/L
• ALT: 38 U/L
• platelets: 140 × 10⁹/L

None of those numbers screams liver failure.

But together with his age they produce a FIB-4 of approximately 2.2.

That does not prove advanced fibrosis.

It does something more appropriate for a screening tool: it says that Robert should not simply be reassured by his aminotransferases.

He undergoes a second-line noninvasive fibrosis assessment with elastography.

The liver stiffness is elevated enough to justify hepatology evaluation and a more deliberate fibrosis workup.

Why this is such an important paradigm

FIB-4 demonstrates that diagnostic progress does not always require inventing a new biomarker.

Sometimes the unused information is already in the EHR.

We merely fail to combine it.

That is a profound point.

An algorithm can turn four ordinary measurements into a new clinical variable.

Modern AASLD guidance uses FIB-4 as a first-line, high-negative-predictive-value strategy to identify people who are unlikely to have advanced fibrosis and to identify those who need secondary assessment such as vibration-controlled transient elastography, magnetic resonance elastography, or the Enhanced Liver Fibrosis test.

Cutoffs are context-dependent, and performance varies with age and population, so the number should not be treated as a universal diagnostic threshold.

Evidence anchor: See the American Association for the Study of Liver Diseases discussion of noninvasive assessment in MASLD.

---

7. Fecal calprotectin: measuring intestinal inflammation instead of guessing from symptoms

Inflammatory bowel disease creates another version of the same problem.

Symptoms are important.

But symptoms are not identical to inflammation.

A patient with Crohn’s disease may feel well while intestinal inflammation continues. Another may have pain, bloating, or diarrhea despite little active inflammatory disease.

Fecal calprotectin is a stool biomarker associated with neutrophilic intestinal inflammation.

It does not replace colonoscopy, imaging, pathology, or clinical judgment.

But it can help answer a question symptoms alone cannot answer reliably:

Is intestinal inflammation probably active right now?

The American Gastroenterological Association recommends combining biomarkers with symptoms rather than relying on symptoms alone in Crohn’s disease management.

Case study: remission by conversation, not by biology

Patient: Priya, 31

Priya has established Crohn’s disease.

At follow-up she says she feels “almost completely normal.”

She has one or two formed bowel movements per day, no fever, no visible blood, and minimal abdominal discomfort.

A symptom-only approach would call this remission.

Her fecal calprotectin, however, has risen from 110 μg/g to 720 μg/g.

The test is repeated to reduce the chance that a transient confounder is driving the result. It remains markedly elevated.

Further evaluation shows active intestinal inflammation.

Her symptoms were telling the truth about how she felt.

The biomarker was telling the truth about a different variable.

Both mattered.

Now reverse the case.

Suppose Priya has persistent abdominal discomfort but repeatedly low inflammatory biomarkers and recent objective confirmation of mucosal remission.

In that situation, automatically escalating immunosuppression because “the Crohn’s must be active” could be exactly the wrong response. Other explanations for symptoms deserve investigation.

This is a recurring theme in laboratory medicine:

A biomarker is most valuable when it measures something different from what the history already tells us.

Evidence anchor: The AGA guideline on biomarkers in Crohn’s disease recommends combining biomarkers and symptoms and describes circumstances in which fecal calprotectin and CRP can help rule out active inflammation.

---

8. Islet autoantibodies and C-peptide: “adult diabetes” is not synonymous with type 2 diabetes

Diabetes is often classified from phenotype.

An adult with hyperglycemia—especially an adult with overweight or obesity—may be assumed to have type 2 diabetes.

That assumption will often be correct.

But “often” is not the same as “always.”

Autoimmune type 1 diabetes can begin in adulthood. Obesity does not immunize someone against autoimmune beta-cell destruction. Some adults lose insulin production gradually enough that they initially resemble type 2 diabetes.

The key laboratory tools ask two different questions:

Islet autoantibodies

Examples include:

• GAD antibodies
• IA-2 antibodies
• ZnT8 antibodies
• insulin autoantibodies in selected circumstances

These ask:

Is there evidence of an autoimmune process directed against pancreatic beta-cell biology?

C-peptide

C-peptide is released when endogenous proinsulin is cleaved into insulin and C-peptide.

It therefore helps answer:

How much insulin is this patient’s own pancreas still producing?

The 2026 ADA Standards of Care recommend standardized islet-autoantibody testing for classification when adults have phenotypic features overlapping type 1 and type 2 diabetes.

Case study: “type 2” diabetes that keeps behaving strangely

Patient: Nia, 39

Nia is diagnosed with diabetes after an A1C of 9.4%.

Her BMI is 30 kg/m².

Because she is an adult with overweight, the initial working diagnosis is type 2 diabetes.

Over the next year, however:

• she loses weight unintentionally
• glycemic control deteriorates rapidly
• several noninsulin therapies produce little durable effect
• insulin becomes necessary much sooner than expected

GAD antibody testing is strongly positive.

A properly timed C-peptide measurement, interpreted together with the simultaneous glucose concentration, demonstrates limited endogenous insulin secretion.

The diagnosis changes from a purely insulin-resistant model to autoimmune beta-cell failure.

That matters because the anticipated disease trajectory, education, insulin strategy, ketone awareness, DKA risk, and screening for associated autoimmune disease all change.

The lesson is not “test every adult with diabetes for everything.”

It is:

When the disease behaves differently from the model, test the model.

The ADA also now recognizes presymptomatic stages of type 1 diabetes in people with islet autoimmunity, particularly those with multiple confirmed autoantibodies and elevated familial or genetic risk. That is a major conceptual shift: laboratory markers can identify autoimmune disease before classical symptomatic diabetes appears.

Evidence anchor: See the ADA Standards of Care in Diabetes—2026: Diagnosis and Classification.

---

9. Pharmacogenomics: useful genetics looks less glamorous than genetic fortune-telling

Genetic testing is one of the easiest areas in medicine to overhype.

A genome contains enormous amounts of information, but enormous information content does not mean every variant has actionable clinical meaning.

The strongest pharmacogenomic use cases are narrower and more disciplined.

They ask:

Does this inherited variant predict a meaningful difference in how this patient metabolizes, activates, transports, or responds to a particular drug?

The Clinical Pharmacogenetics Implementation Consortium, or CPIC, publishes evidence-graded gene-drug guidelines for situations in which genotype can inform prescribing.

Examples include:

• DPYD and fluoropyrimidines
• TPMT / NUDT15 and thiopurines
• CYP2C19 and several drug classes
• CYP2D6 and selected medications
• RYR1 / CACNA1S and malignant-hyperthermia-triggering anesthetics

CPIC makes an important distinction: its guidelines primarily answer how to use a genotype if it is available, not necessarily who must be tested.

That restraint is important.

Case study: the laboratory result that matters before the first dose

Patient: Samuel, 62

Samuel is about to begin a fluoropyrimidine-based chemotherapy regimen.

A clinically validated DPYD genotype is available before treatment.

The result predicts reduced dihydropyrimidine dehydrogenase activity.

Without that information, Samuel could receive a conventional starting regimen despite having impaired ability to metabolize the drug normally, increasing the risk of severe toxicity.

With the information, the oncology team can use an established pharmacogenomic dosing framework and intensified monitoring appropriate to the phenotype.

This is a fundamentally different use of genetics from ordering a wellness genome panel and trying to infer the ideal vitamin stack.

The chain of reasoning is tight:

$$\text{variant} \rightarrow \text{enzyme function} \rightarrow \text{drug exposure} \rightarrow \text{toxicity/efficacy risk} \rightarrow \text{prescribing decision}$$

That is what actionable molecular medicine looks like.

Evidence anchors: CPIC maintains evidence-based guidance for DPYD and fluoropyrimidines and TPMT/NUDT15 and thiopurines. CPIC also explicitly states that its guidelines are generally designed to explain how available genetic results should be used rather than whether a test must be ordered; see CPIC Guidelines.

---

The deeper problem: medicine underuses paradigms, not just tests

If we stopped here, the solution would appear to be a longer laboratory requisition.

That is not the solution.

The deeper opportunity is to improve the logic connecting tests together.

Several paradigms deserve much more attention.

---

Paradigm 1: compare patients with themselves, not only with populations

Reference intervals are usually derived from populations.

A typical reference interval might contain approximately the central 95% of results from a defined reference population under specified conditions.

That makes reference intervals useful.

It also creates a dangerous psychological shortcut:

Inside interval = normal. Outside interval = abnormal.

But a patient is not a population.

Case study: the creatinine that never turned red

Patient: Luis, 34

Six months ago:

creatinine = 0.72 mg/dL

Today:

creatinine = 1.06 mg/dL

The laboratory reference interval extends beyond 1.06 mg/dL, so the result does not receive a red flag.

A clinician scanning only the flags might move on.

But Luis’s creatinine has increased by nearly 50% from his previous baseline.

The absolute value and the trajectory tell different stories.

The correct question is not only:

Is 1.06 outside the population reference interval?

It is also:

Why did this particular patient’s value change this much?

The same logic applies to:

• hemoglobin
• platelet count
• liver enzymes
• sodium
• troponin
• PSA
• eGFR
• tumor markers when appropriately used for monitoring
• inflammatory markers

Sometimes the delta carries more information than the absolute position.

Clinical laboratories already understand biological variation and analytical imprecision. Laboratory information systems and EHRs could, in principle, make personalized trajectories far more visible than they usually are.

Evidence anchor: CLSI emphasizes that reference intervals are population- and method-dependent constructs rather than magical borders between health and disease. See CLSI EP28: Defining, Establishing, and Verifying Reference Intervals.

---

Paradigm 2: distinguish a reference interval from a medical decision limit

These concepts are related but not identical.

A reference interval describes the distribution of results in a selected reference population.

A medical decision limit is a threshold chosen because evidence connects it with a clinical action, risk classification, diagnosis, or management decision.

Confusing them causes trouble.

Consider hemoglobin A1C.

The diagnostic thresholds for diabetes are not simply the upper edge of whatever A1C distribution happens to exist in a healthy local reference population. They are clinically defined decision thresholds supported by outcome data and diagnostic standards.

Likewise, cardiac troponin interpretation depends on assay-specific thresholds, clinical context, serial change, symptoms, and evidence of ischemia—not merely a generic HIGH flag.

This distinction matters because the laboratory report visually presents everything as if it were the same kind of boundary.

It is not.

---

Paradigm 3: use physiology to predict the companion abnormality

This is one of the most powerful habits in clinical reasoning.

Do not stop at:

What diseases cause this abnormal result?

Ask:

If my hypothesis is correct, what else should I expect to see?

Examples:

Hypothesis: primary aldosteronism

Predict:

• suppressed renin
• aldosterone inappropriately high relative to renin
• hypertension
• sometimes hypokalemia

Hypothesis: iron-deficiency anemia

Predict:

• low hemoglobin
• often low MCV
• low ferritin in uncomplicated deficiency
• low transferrin saturation
• often increased iron-binding capacity
• compatible clinical source of iron loss or increased requirement

Hypothesis: intravascular hemolysis

Predict some combination of:

• anemia
• reticulocytosis if marrow response is intact
• elevated LDH
• elevated indirect bilirubin
• decreased haptoglobin
• characteristic smear findings depending on mechanism

Hypothesis: glomerular kidney injury

Predict some combination of:

• albuminuria or proteinuria
• hematuria in some diseases
• abnormal urine sediment in some diseases
• changing eGFR depending on severity and stage

Hypothesis: cholestatic hepatobiliary disease

Predict a pattern in which alkaline phosphatase is disproportionately elevated relative to aminotransferases, with bilirubin and imaging adding context.

The point is not to memorize these as diagnostic equations.

The point is to use one result to make testable predictions about the rest of the system.

That turns laboratory interpretation into falsifiable reasoning.

---

Paradigm 4: use orthogonal measurements

A laboratory result becomes far more convincing when a different methodology independently supports the same physiological story.

Examples:

FIB-4 → elastography

Blood values suggest fibrosis risk.

A mechanical measurement asks whether the liver is actually stiff.

BNP/NT-proBNP → echocardiography

A peptide concentration suggests hemodynamic cardiac stress.

Ultrasound directly evaluates cardiac structure and function.

Troponin → ECG + imaging + serial kinetics

A protein released during myocardial injury is interpreted alongside electrical patterns, symptoms, and sometimes coronary or cardiac imaging.

Albuminuria → urine microscopy + serology + biopsy when indicated

A quantitative leak suggests glomerular disease.

Microscopy may reveal a cellular pattern. Serologies may identify an immune mechanism. Histology can reveal actual tissue architecture.

Monoclonal protein → electrophoresis + immunofixation + free light chains + marrow/tissue evaluation when indicated

Multiple laboratory techniques interrogate different properties of the same suspected clonal process.

This is a powerful general principle:

Corroboration is strongest when the second test fails in a different way than the first.

Two repetitions of the same assay reduce random error.

Two different modalities can reduce conceptual error.

---

Paradigm 5: design laboratory cascades instead of ordering flat mega-panels

A flat mega-panel asks dozens or hundreds of questions simultaneously whether or not the first answers justify the next ones.

A diagnostic cascade is more disciplined.

It asks one high-value question and allows that answer to determine the next test.

For example:

$$\text{abnormal TSH} \rightarrow \text{free T4 in the appropriate context}$$

or:

$$\text{elevated FIB\text{-}4} \rightarrow \text{elastography / ELF / further fibrosis assessment}$$

or:

$$\text{positive aldosterone-renin screen} \rightarrow \text{repeat / confirmatory strategy as appropriate} \rightarrow \text{subtype evaluation if treatment decisions require it}$$

or:

$$\text{unexplained monoclonal-protein signal} \rightarrow \text{immunofixation / free light chains} \rightarrow \text{hematologic evaluation}$$

This approach minimizes unnecessary testing while allowing very deep testing when the physiology earns it.

The ideal diagnostic system is therefore not:

Order everything.

It is:

Order the next test that most efficiently distinguishes between the leading competing explanations.

---

Paradigm 6: optimize tests for the question they are good at answering

Not every test is equally good at confirming and excluding disease.

Some are especially useful because a low-risk result has a high negative predictive value in the right population.

FIB-4 is often used this way in metabolic liver disease: a low-risk result can help identify people unlikely to have advanced fibrosis, while an elevated result is not itself a diagnosis and instead triggers more specific testing.

D-dimer is another classic example in a different domain. In an appropriately selected low- or intermediate-pretest-probability patient, a negative result can help exclude venous thromboembolism. In a very high-pretest-probability patient, the testing strategy changes. In a hospitalized inflammatory patient, a positive D-dimer may be extremely nonspecific.

The same number has different diagnostic value in different populations because:

$$\text{post-test probability} = f(\text{test performance}, \text{pretest probability})$$

That is Bayes’ theorem hiding underneath laboratory medicine.

A test result does not carry its clinical meaning around independently of the patient.

---

Paradigm 7: sometimes the missing test is more informative than an abnormal test

Imagine reviewing a chart with:

• diabetes for 12 years
• hypertension
• annual CMPs
• serial creatinine values
• serial eGFR values

But no urine albumin measurement in five years.

The problem is not an abnormal laboratory result.

The problem is missing information in a domain known to matter.

Likewise:

• a strong family history of premature coronary disease with no Lp(a) ever measured
• years of hypertension with no aldosterone-renin assessment despite an appropriate indication
• metabolic liver risk with serial aminotransferases but no fibrosis risk stratification
• rapidly progressive “type 2 diabetes” without reconsideration of classification

A high-quality diagnostic system should therefore be capable of asking:

Given what we already know about this patient, which important physiologic dimension has never actually been measured?

That is a much more interesting use of clinical decision support than merely coloring abnormal results red.

---

What about MTHFR, food sensitivity panels, micronutrient mega-panels, and other popular tests?

This is where the distinction between underused evidence-based testing and overmarketed testing becomes essential.

A test can be technically measurable without being clinically useful.

A genetic variant can be real without being a useful explanation for a patient’s symptoms.

An antibody can be detectable without proving intolerance.

A biomarker can correlate with disease without improving decisions.

The right questions are:

1. Does the test measure what it claims to measure?
2. Is the association with the clinical condition reproducible?
3. Does the result add information beyond what we already know?
4. Does it alter management?
5. Does acting on it improve meaningful outcomes?
6. What is the false-positive burden?
7. Is the interpretation validated in the population being tested?

This is why a once-per-adulthood Lp(a) measurement and an indiscriminate “cardiometabolic genetics optimization panel” are not equivalent simply because both sound advanced.

Likewise, common MTHFR polymorphisms are biochemically real but are frequently asked to carry much more diagnostic meaning than evidence supports.

The scientific frontier is usually less flashy than the marketing frontier.

---

A larger case study: what systems-based laboratory medicine could look like

Consider a fictional patient named Daniel.

Daniel is 52. He has obesity, hypertension, type 2 diabetes, and a father who died from a myocardial infarction at 54.

At his annual visit, a traditional routine panel produces:

CBC

Unremarkable.

CMP

• creatinine: 0.91 mg/dL
• eGFR: 96 mL/min/1.73 m²
• AST: 35 U/L
• ALT: 42 U/L
• potassium: 3.9 mmol/L

A1C

7.4%

Lipid panel

• LDL-C: 93 mg/dL
• HDL-C: 41 mg/dL
• triglycerides: 228 mg/dL

A superficial interpretation might be:

Diabetes could be better controlled. Triglycerides are high. Kidney function looks good. Liver enzymes are not very concerning. LDL is under 100. Continue routine follow-up.

Now consider a systems-based approach.

Cardiovascular system

Because Lp(a) has never been measured, it is checked once.

Lp(a): 210 nmol/L

ApoB is also measured because of diabetes and hypertriglyceridemia.

ApoB: 112 mg/dL

His apparently reassuring LDL-C is now understood in the context of a high inherited Lp(a) burden and increased atherogenic particle number.

Kidney system

Creatinine looks good, but UACR is measured.

UACR: 240 mg/g

Persistent albuminuria is confirmed.

Daniel has meaningful kidney involvement despite preserved filtration.

Liver system

His aminotransferases do not look dramatic, but his metabolic risk is high.

Age, AST, ALT, and platelets are used to calculate FIB-4.

The result falls into a range that warrants secondary fibrosis assessment.

Elastography shows increased liver stiffness.

Endocrine hypertension system

Daniel’s hypertension has required several medications over time.

Aldosterone and renin are checked under an appropriate testing protocol.

Renin is suppressed and aldosterone is inappropriately high relative to renin.

Repeat evaluation supports further workup for primary aldosteronism.

The point is not that Daniel had four hidden diseases waiting to be discovered in every patient.

The point is that the ordinary panel sampled four systems incompletely.

The extra information was not random.

Each test answered a question already justified by Daniel’s phenotype.

That is the model worth pursuing:

$$\text{phenotype} \rightarrow \text{specific unanswered physiologic question} \rightarrow \text{targeted measurement} \rightarrow \text{updated probability} \rightarrow \text{next decision}$$

Not:

$$\text{patient} \rightarrow \text{order 200 biomarkers} \rightarrow \text{circle everything outside the range}$$

---

Why does the implementation gap exist?

If these tests are known and useful, why are they not uniformly used?

There is no single reason.

1. Guidelines change faster than workflows

A physician may have trained when primary aldosteronism screening was reserved for selected high-risk hypertension. The 2025 Endocrine Society guideline now suggests screening all people with hypertension.

The science can change overnight on paper.

The EHR order set, insurance policy, clinic habit, and clinician mental model may take years to follow.

2. Routine panels are convenient

CBC and CMP order buttons are effortless.

Lp(a), ApoB, UACR, cystatin C, autoantibodies, or specialized stool markers require the clinician to remember the specific question.

Healthcare systems naturally optimize for repeatable workflows.

That is good for reliability but bad for edge cases and newly recognized risk dimensions.

3. Specialists and primary care live in different informational worlds

A nephrologist thinks naturally in eGFR and albuminuria categories.

An endocrinologist may think naturally in renin-aldosterone physiology.

A lipid specialist may think in ApoB particle number and Lp(a).

A gastroenterologist may think in objective inflammatory biomarkers plus symptoms.

Primary care must cover all of those worlds simultaneously.

The implementation problem is partly an information-architecture problem.

4. Test interpretation can be harder than test ordering

Anyone can order renin and aldosterone.

Correctly interpreting them in the setting of medications, potassium status, sodium intake, posture, assay type, kidney disease, and pretest probability is harder.

Anyone can calculate FIB-4.

Understanding its population limitations and knowing the appropriate next test is harder.

The bottleneck is often not laboratory capability.

It is decision support.

5. Insurance and access are uneven

A guideline recommendation does not guarantee affordable access to every laboratory assay, imaging modality, specialist, or confirmatory procedure.

The theoretically optimal diagnostic pathway may be difficult to execute in a rural clinic, underinsured population, or fragmented health system.

6. Medicine correctly fears overtesting

Modern healthcare already suffers from incidental findings and low-value testing.

Clinicians are rightly cautious about opening diagnostic cascades.

The answer is therefore not indiscriminate expansion.

It is more selective sophistication.

---

What an ideal laboratory report might eventually look like

Imagine opening a laboratory report that does more than color numbers red.

It might show:

Population context

“Creatinine is within the laboratory reference interval.”

Personal context

“Creatinine has increased 42% from the patient’s median value over the previous two years.”

Related-system context

“eGFR has declined concurrently. No UACR has been measured in the previous 18 months.”

Analytical context

“Specimen hemolysis may falsely increase potassium.”

Conditional context

“Because this patient has diabetes, current kidney guidelines recommend assessment of albuminuria in addition to filtration.”

Hypothesis context

“If the elevated alkaline phosphatase is hepatic in origin, consider corroboration with GGT or other clinically appropriate evaluation; if nonhepatic, consider bone sources.”

Longitudinal visualization

A graph showing the last five years rather than a single disconnected number.

Decision support

Not a diagnosis generated by an algorithm, but a reminder of which physiologic question remains unanswered.

This is an area where software, laboratory medicine, and artificial intelligence could eventually intersect extremely productively.

The laboratory already generates enormous amounts of structured quantitative data.

What is often missing is the layer that turns measurements into relationships, trajectories, and hypotheses.

---

A compact framework for using uncommon tests intelligently

Before ordering an unfamiliar laboratory test, ask six questions.

1. What physiological variable does it measure?

Not “what diseases is it associated with?”

What does it actually measure?

Lp(a) measures a specific lipoprotein burden.

UACR measures albumin leakage relative to urinary creatinine.

C-peptide reflects endogenous insulin secretion.

Fecal calprotectin reflects intestinal inflammatory activity.

2. Why is the ordinary test insufficient here?

If LDL-C already answers the question, ApoB may add little.

If creatinine is reliable and the clinical decision does not require greater GFR accuracy, cystatin C may add little.

If the disease phenotype fits perfectly, an exotic test may be unnecessary.

3. What would I do differently if the result were high?

If the answer is “nothing,” reconsider ordering it.

4. What would I do differently if the result were low or negative?

Tests are frequently most useful because they allow clinicians to safely stop a diagnostic pathway.

5. What can make the result misleading?

Think about:

• medications
• fasting state
• posture
• exercise
• acute illness
• pregnancy
• age
• muscle mass
• kidney function
• inflammation
• specimen handling
• assay methodology
• pretest probability

6. What independent observation could confirm or refute the story?

A different biomarker?

A repeat measurement?

Imaging?

Pathology?

Physiologic testing?

A response to a targeted intervention?

That final question turns a laboratory value into part of an investigation rather than an endpoint.

---

The tests in one table

Test / paradigm	What the ordinary workflow may miss	What the additional information can reveal
Lp(a)	Inherited cardiovascular risk outside conventional lipid values	Genetically driven atherogenic risk that may justify more aggressive prevention
ApoB	LDL-C can underestimate particle number	Total burden of atherogenic particles, especially in metabolic disease or high triglycerides
Aldosterone + renin / ARR	Hypertension is often treated as a single disease	Autonomous aldosterone physiology and a potentially specifically treatable cause
UACR	Creatinine/eGFR measure filtration but not glomerular albumin leak	Kidney damage and cardiovascular risk despite preserved filtration
Cystatin C	Creatinine is distorted by unusual muscle mass and other non-GFR factors	An independent filtration estimate and improved combined eGFR in selected patients
FIB-4	Normal or mildly abnormal AST/ALT can coexist with fibrosis	Advanced-fibrosis risk using ordinary existing laboratory data
Fecal calprotectin	Symptoms and intestinal inflammation can diverge	Objective evidence for or against active inflammatory disease
Islet autoantibodies	Adult diabetes may be assumed to be type 2	Autoimmune beta-cell destruction
C-peptide	Glucose and A1C do not reveal endogenous insulin reserve	How much insulin the patient’s pancreas is producing
Pharmacogenomics	Standard drug dosing assumes average metabolism	Genetically altered drug metabolism or response for selected gene-drug pairs
Longitudinal change	A result may remain inside the population reference interval	A clinically meaningful deviation from the patient’s own baseline
Orthogonal confirmation	One modality can mislead in characteristic ways	Independent evidence from a method with different assumptions and failure modes

---

The most important conclusion is not “order more tests”

It is tempting to end an article like this with a giant list titled:

Twenty laboratory tests to demand from your doctor.

That would betray the entire point.

The lesson is not that every person should receive every test described here.

The lesson is that the familiar laboratory panels are sampling instruments, not complete maps of human physiology.

They answer particular questions.

When the clinical situation raises a different question, medicine already possesses many validated tools for looking deeper.

The intellectual shift is from this:

Is the CBC normal?

Is the CMP normal?

Is the lipid panel normal?

To this:

What physiological system am I trying to understand?

Which dimensions of that system have already been measured?

Which important dimension remains unmeasured?

If my hypothesis is true, what else should I observe?

What test would most efficiently distinguish between the competing explanations?

That is not alternative medicine.

It is laboratory medicine taken seriously.

And perhaps the most underused laboratory technology of all is not a biomarker.

It is reasoning across biomarkers.

---

Sources and further reading

The recommendations and examples in this article are grounded primarily in current professional guidelines and laboratory standards. Clinical decisions should use the complete guideline, local laboratory methods, and individual patient context rather than this overview alone.

• American College of Cardiology / American Heart Association. 2026 Guideline on the Management of Dyslipidemia. Summary: ACC/AHA Issue Updated Guideline for Managing Lipids, Cholesterol.
• Endocrine Society. Primary Aldosteronism: An Endocrine Society Clinical Practice Guideline. July 2025: Primary Aldosteronism.
• Kidney Disease: Improving Global Outcomes. KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease: CKD Evaluation and Management.
• American Association for the Study of Liver Diseases. Noninvasive Assessment of Patients with MASLD: Spare Me the Jab.
• American Gastroenterological Association. The Role of Biomarkers for the Management of Crohn’s Disease: AGA Clinical Guidance.
• American Diabetes Association. Standards of Care in Diabetes—2026: Diagnosis and Classification of Diabetes: Diabetes Care.
• Clinical Pharmacogenetics Implementation Consortium. CPIC Guidelines: CPIC.
• Clinical and Laboratory Standards Institute. EP28: Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory: CLSI EP28.
• Clinical and Laboratory Standards Institute. EP29: Expression of Measurement Uncertainty in Laboratory Medicine: CLSI EP29.