What Baseline Blood Work Measures, and How to Store It So It Compounds

There is no blood test for everything. What people mean by “baseline blood work” is a bundle of four or five standard panels, roughly 40 to 60 analytes.

• a complete blood count with differential
• a comprehensive metabolic panel
• a lipid panel
• thyroid function
• iron studies
• a glycemic and inflammatory add-on

That bundle is cheap and fast. It mostly measures kidney function, liver enzymes and electrolytes, along with red and white cell populations, lipid transport, and glucose handling. Its value is not in the first draw. A single result tells you where you sit inside a population reference interval, which is a weak statement. Three or four draws under matched conditions tell you your own set point and the direction you are moving, which is the thing worth having.

What a “full panel” means on a requisition

Order codes vary by lab, but the standard stack is:

• CBC with differential: hemoglobin, hematocrit, MCV, RDW and platelets, plus absolute neutrophil / lymphocyte / monocyte / eosinophil counts.
• CMP (14 analytes) covers the following.
  • sodium
  • potassium
  • chloride
  • CO2
  • BUN
  • creatinine
  • glucose
  • calcium
  • total protein
  • albumin
  • total bilirubin
  • ALP
  • AST
  • ALT
• Lipid panel: total cholesterol, HDL-C, triglycerides and LDL-C. LDL-C is usually calculated by Friedewald or Martin-Hopkins. It is not measured.
• TSH, with free T4 reflexed or ordered outright.
• Ferritin, serum iron, TIBC and transferrin saturation.

To that we would add the following on a first draw.

• apolipoprotein B
• lipoprotein(a) once in a lifetime (it is ~90% genetically determined and does not need repeating)
• HbA1c
• fasting insulin
• hs-CRP
• GGT
• uric acid
• 25-hydroxyvitamin D
• homocysteine
• a urinalysis with albumin-to-creatinine ratio

ApoB is the one we would not skip: it counts atherogenic particles directly instead of estimating the cholesterol they carry, and the two diverge in exactly the people whose triglycerides are high and LDL-C looks unremarkable.

Do not order a broad tumor marker panel as a screening exercise. CA-125, CEA, and CA 19-9 have low positive predictive value in unselected people and reliably generate imaging cascades. Multi-analyte approaches do better when combined with imaging and a defined follow-up path. A prospective study of 10,006 women with no history of cancer used a blood test combining protein markers with circulating tumor DNA, followed by PET-CT in test-positive participants. It detected 26 cancers that were found by the blood test first, at a specificity of 99.6%.¹ That is a research protocol with a full workup attached. It is not something to replicate with a la carte ordering.

The pre-analytical variables that corrupt a longitudinal series

Most of the variance people attribute to “my health changed” is collection technique. Control these or your trend lines are noise:

• Time of day. Cortisol, testosterone, and iron all swing severalfold across 24 hours. Serum iron can vary by 30% or more between morning and evening. Always draw at the same hour, ideally between 07:00 and 09:00.
• Fasting state. Triglycerides, glucose, and insulin are the sensitive ones. Fast 10 to 12 hours, water only.
• Posture. Standing shifts water out of the vascular compartment and concentrates large molecules. Albumin, total protein, hemoglobin, and lipids read roughly 5 to 10% higher after standing versus supine. Sit quietly for 5 minutes before the draw, every time.
• Tourniquet time. Over a minute of stasis inflates potassium and protein-bound analytes. Ask for release once blood flows.
• Recent exercise. A hard session raises CK, AST, and ALT and can keep them elevated for days. Leave 48 hours.
• Hemolysis. Hemolyzed samples falsely elevate potassium, LDH, and AST. Labs report a hemolysis index. Ask for it and discard affected results rather than interpreting them.
• Vendor and platform. Ferritin, TSH, and free T4 assays are not harmonized across manufacturers. Switching labs mid-series introduces a step change you will misread as biology. Pick one lab and stay.

Get the data as structured records and not as PDFs

Lab portals hand you PDFs. Parsing them is a losing game: layouts change, reference intervals move, and units are inconsistent. The better path is the FHIR R4 API that most US labs and health systems expose under patient access rules. Pull Observation and DiagnosticReport resources and keep the raw JSON forever.

What matters in each Observation:

• code.coding with a LOINC identifier. Key on LOINC. Do not key on the display string. “Glucose”, “Glucose, Serum”, and “GLU” are the same thing, while 2345-7 is stable. Creatinine is 2160-0, ALT is 1742-6, HbA1c is 4548-4.
• valueQuantity.unit plus valueQuantity.code in UCUM. Normalize on ingest. Ferritin in ng/mL and µg/L are numerically identical, HbA1c in % (NGSP) and mmol/mol (IFCC) are not.
• referenceRange, stored per result, because it is a property of the lab and the assay. It is not a property of you.
• effectiveDateTime, and store the collection time. Do not store the report time.

A flat table of (loinc, value, unit_ucum, collected_at, lab_id, method, hemolysis_index) is enough to do everything below. Add a fasting_hours column you fill in yourself.

Deciding whether a change is real

Two results differ. The question is whether the difference exceeds what measurement noise and your own day-to-day biology would produce anyway. The reference change value answers it:

import math

def rcv(cv_analytical, cv_intra, z=1.96):
    """Percent change needed for significance at ~95%, two-sided."""
    return z * math.sqrt(2) * math.sqrt(cv_analytical**2 + cv_intra**2)

rcv(2.0, 1.7)   # HbA1c  -> ~7.3%
rcv(3.0, 9.0)   # LDL-C  -> ~26%
rcv(5.0, 40.0)  # hs-CRP -> ~112%

Within-subject biological variation (CV_I) is published per analyte in the EFLM Biological Variation Database. The spread is enormous. Sodium sits near 0.6%, HbA1c near 2% and creatinine around 5%. ALT sits in the high teens to low twenties, ferritin near 15% and hs-CRP well above 40%. The practical consequence: an hs-CRP that goes from 0.8 to 1.6 mg/L means essentially nothing on its own, while an HbA1c moving from 5.2 to 5.6% is a real signal. Compute RCV per analyte once, store it next to the LOINC code, and let your tooling flag only crossings.

Second statistical point: reference intervals are central 95% ranges from a reference population. Run 50 independent analytes on a healthy person and the chance of at least one out-of-range flag is about 1 − 0.95^50 ≈ 92%. A single asterisk on a report is the expected outcome. It is not a finding. Patterns across correlated analytes are what carry information. Large cohort work shows routine blood markers cluster into shared signals that track multiple chronic conditions at once rather than mapping one marker to one disease.²

What needs a clinician, immediately

Some results are not trend material. Take them to a physician the same day.

• hemoglobin below roughly 8 g/dL or any rapid drop
• platelets below 50 ×10⁹/L
• potassium below 3.0 or above 6.0 mmol/L
• absolute neutrophil count below 0.5 ×10⁹/L
• a corrected calcium above 12 mg/dL
• transaminases above 10× the upper limit of normal
• a creatinine that has risen more than 50% from your own baseline

Blasts or unexplained circulating immature cells on a differential are the same category. We are describing measurement. We are not describing care. Interpretation that leads to treatment belongs with a licensed clinician who can see you.

Routine counts carry more prognostic weight than their price suggests. In advanced melanoma patients starting immunotherapy, baseline absolute lymphocyte count, eosinophil count, and LDH drawn before treatment were associated with survival, all from a standard CBC and chemistry panel.³

What blood work does not see

A standard panel is a snapshot of plasma chemistry. It does not see coronary plaque burden, most early solid tumors, or structural disease. Protein panels can add resolution beyond routine chemistry, in colorectal cancer detection⁴ and in neurodegeneration, where plasma p-tau and neurofilament light now predict dementia risk in clinic populations and track cognitive decline in community cohorts.⁵⁶ But panel performance drops sharply when you move from case-control design to unselected screening, which a systematic review of blood-based panels for cognitive impairment makes plain.⁷ Treat any panel’s headline AUC as an upper bound that will not survive contact with your base rate.

Questions people also ask

What is a full panel blood test called? There is no single name. Labs sell bundles under marketing names, but the clinical contents are a CBC with differential, a comprehensive metabolic panel, and a lipid panel, sometimes with TSH and iron studies. Ask for the LOINC list. Do not settle for the bundle name.

What are red flags on a blood test? Severe cytopenias, potassium outside 3.0 to 6.0 mmol/L, or calcium above 12 mg/dL. Others are transaminases above 10× the upper limit, a sharp creatinine rise, or abnormal cells on a differential. These warrant same-day clinical review. They do not call for self-interpretation.

Is there a blood test that checks for everything? No. Multi-analyte tests exist and are improving, but they are evaluated inside protocols that pair them with imaging and defined follow-up.¹ Nothing on the consumer menu screens for all disease.

What can a standard blood test reveal? It reveals anemia and its likely class via MCV and RDW, kidney filtration via creatinine and eGFR, and liver injury patterns via ALT/AST/ALP/GGT. It also shows thyroid state, iron stores, and glucose handling over ~3 months via HbA1c. The last items are atherogenic particle burden via ApoB and a coarse read on systemic inflammation.

How often should I repeat it? Annually if stable, and quarterly for the first year if you want to establish a personal set point with tight enough error bars to detect real drift.

Footnotes

1. Anne Marie Lennon, Adam H. Buchanan, Isaac Kinde, et al. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science, 2020. https://doi.org/10.1126/science.abb9601 ↩ ↩²

2. Alice Margherita Ornago, Caterina Gregorio, Federico Triolo, et al. Shared and specific blood biomarkers for multimorbidity. Nature Medicine, 2026. https://doi.org/10.1038/s41591-025-04038-2 ↩

3. Alexander Martens, Kilian Wistuba‐Hamprecht, Marnix Geukes Foppen, et al. Baseline Peripheral Blood Biomarkers Associated with Clinical Outcome of Advanced Melanoma Patients Treated with Ipilimumab. Clinical Cancer Research, 2016. https://doi.org/10.1158/1078-0432.ccr-15-2412 ↩

4. Kim Y. C. Fung, Bruce Tabor, Michael Buckley, et al. Blood-Based Protein Biomarker Panel for the Detection of Colorectal Cancer. PLoS ONE, 2015. https://doi.org/10.1371/journal.pone.0120425 ↩

5. Kumar B. Rajan, Neelum T. Aggarwal, Elizabeth A. McAninch, et al. Remote Blood Biomarkers of Longitudinal Cognitive Outcomes in a Population Study. Annals of Neurology, 2020. https://doi.org/10.1002/ana.25874 ↩

6. Vincent Planche, Vincent Bouteloup, Isabelle Pellegrin, et al. Validity and Performance of Blood Biomarkers for Alzheimer Disease to Predict Dementia Risk in a Large Clinic-Based Cohort. Neurology, 2022. https://doi.org/10.1212/wnl.0000000000201479 ↩

7. Anette Hardy-Sosa, Karen León‐Arcia, Jorge J. Llibre‐Guerra, et al. Diagnostic Accuracy of Blood-Based Biomarker Panels: A Systematic Review. Frontiers in Aging Neuroscience, 2022. https://doi.org/10.3389/fnagi.2022.683689 ↩