What a Comprehensive Blood Panel Measures, and How to Read Your Own

A comprehensive blood panel, as sold by most labs, means a comprehensive metabolic panel (CMP) plus a complete blood count (CBC). It usually adds a lipid panel and sometimes thyroid-stimulating hormone (TSH). That is 14 analytes covering electrolytes, kidney and liver chemistry, and glucose. It also includes about 10 hematology indices, four lipid measures, and one hormone. It is a reasonable structural survey of kidney, liver, and blood-cell function. On its own it is a weak metabolic and cardiovascular assessment. If you order a “comprehensive” panel expecting a full picture, the gaps are in the parts that carry the most information about long-run risk. A CMP does not include the following tests.

• hemoglobin A1c (HbA1c)
• insulin
• apolipoprotein B (apoB)
• lipoprotein(a)
• high-sensitivity C-reactive protein (hs-CRP)
• ferritin
• vitamin D

Below is what we would order, how to draw it so the numbers mean something, and how to decide whether a change between two draws is real.

What belongs in a panel, by tier

The first tier is the standard chemistry and hematology, and it earns its place. The CMP gives the analytes below, along with an estimated glomerular filtration rate (eGFR) derived from creatinine using the race-free 2021 CKD-EPI equation.

• sodium
• potassium
• chloride
• bicarbonate
• glucose
• blood urea nitrogen
• creatinine
• calcium
• total protein
• albumin
• globulin (calculated)
• alanine aminotransferase (ALT)
• aspartate aminotransferase (AST)
• alkaline phosphatase
• total bilirubin

The CBC with differential gives the following.

• hemoglobin
• hematocrit
• red cell count
• mean corpuscular volume
• red cell distribution width
• platelets
• the five white-cell subsets

These are structural: they detect organ dysfunction, anemia, and cytopenias that you would want to know about regardless of what else you are tracking.

The second tier is what we would add and what the marketing panels usually omit. For lipids, order apoB and lipoprotein(a) alongside the standard panel. ApoB counts atherogenic particles directly rather than estimating the cholesterol they carry, which matters because calculated LDL cholesterol and particle count diverge in people with high triglycerides or small dense particles. Insist that calculated LDL use the Martin-Hopkins method rather than Friedewald, which assumes a fixed triglyceride-to-VLDL-cholesterol ratio of 5 and becomes invalid above roughly 400 mg/dL of triglycerides. Lipoprotein(a) is largely genetically determined and needs measuring once, ideally in nmol/L rather than mg/dL, because the mass units depend on isoform size.

For glucose metabolism, add HbA1c and fasting insulin. Add fasting C-peptide if it is available. HbA1c reports a roughly 8 to 12 week average of glycation weighted toward the recent weeks, and it is distorted by anything that changes red cell lifespan, including iron deficiency, hemolysis, recent blood donation, and hemoglobin variants. Fasting insulin with fasting glucose lets you compute HOMA-IR, which is crude but longitudinally informative in one person. For iron, order ferritin, serum iron, transferrin saturation, and total iron-binding capacity together, since ferritin alone confounds storage iron with inflammation. For thyroid, TSH alone is a screening test. Free T4 plus TSH is the interpretable pair. Add 25-hydroxyvitamin D, hs-CRP, and uric acid. Also add gamma-glutamyl transferase and homocysteine. Cystatin C is worth including once because creatinine-based eGFR is sensitive to muscle mass and creatine supplementation, and a large discordance between the two estimates is itself a finding.

Resist the temptation to keep adding. Melander and colleagues tested a set of then-novel cardiovascular biomarkers on top of conventional risk factors in a community cohort. They found only marginal improvement in discrimination, with the C-statistic moving very little.¹ The lesson generalizes. The marginal analyte usually adds variance rather than information, unless you have a specific question it answers.

Preanalytical variables move the numbers more than your biology does

Most surprising lab results are artifacts of collection. The largest single effect is posture. Moving from standing to supine shifts roughly 600 to 700 mL of plasma water into the vasculature over 20 to 30 minutes, which dilutes everything protein-bound by on the order of 5 to 10 percent: albumin, total protein, calcium, hemoglobin, and hematocrit all read lower supine. If you draw seated one visit and supine the next, you have manufactured a trend. Pick one posture and hold it, seated for at least five minutes before the needle.

Tourniquet time is second. Venous stasis beyond about one minute raises potassium, lactate, and protein-bound analytes. Fist clenching during the draw can raise potassium by 1 to 2 mmol/L by itself. Hemolysis from a small-gauge needle or rough handling releases intracellular potassium, lactate dehydrogenase, and AST. Most labs flag it with a hemolysis index. Check for that flag before you believe an isolated high potassium.

Fasting matters selectively. Glucose and insulin require a fast of 8 to 12 hours to be interpretable. Triglycerides rise substantially after a meal, which propagates into any calculated LDL. Non-fasting lipids are acceptable for risk assessment and are what modern guidelines favor, but if you are tracking your own trend, fasting draws are more comparable to each other. Almost nothing else on the panel needs a fast. Iron and vitamin B12 rise sharply after supplements, so hold supplements for 24 hours before an iron panel. Strenuous exercise in the preceding 48 hours raises creatine kinase, AST, ALT, and sometimes creatinine, and heavy resistance training can push ALT well above the reference range in a healthy person. Cortisol, testosterone, and iron all have diurnal rhythms. Standardize the draw time to the morning.

The best thing to eat the day before a blood test is your normal diet. A deliberately clean day distorts triglycerides and glucose away from your actual baseline, which defeats the purpose of measuring. What you should avoid is alcohol within 48 hours, a new supplement regimen, and an unusually hard workout.

Reference ranges are not decision thresholds

A reference interval is the central 95 percent of a nominally healthy population, which means one healthy person in 20 falls outside it for any given analyte by construction. Run 30 analytes and the probability that every one lands inside is about 0.95 to the 30th power, roughly 21 percent. Being flagged on a comprehensive panel is the usual outcome, not the exception, and a single mildly out-of-range value in isolation is mostly a prompt to repeat the measurement rather than to act.

The quantitative tool for longitudinal reading is the reference change value (RCV). Each analyte has a within-subject biological coefficient of variation (CVi), the fluctuation around your own set point, and the lab’s assay has an analytical CV (CVa). The RCV at 95 percent confidence is 1.96 × √2 × √(CVi² + CVa²), about 2.77 times the combined CV. The CVi values differ enormously. HbA1c is tight, around 2 percent, so a change of 0.3 percentage points is likely real. Total cholesterol sits near 6 percent and creatinine near 5 percent. ALT and TSH sit near 20 percent, and hs-CRP near 40 to 60 percent. That means an hs-CRP that moves from 0.8 to 1.9 mg/L tells you almost nothing, while a creatinine moving from 0.95 to 1.15 mg/dL exceeds its RCV and deserves a repeat draw. Compute RCV per analyte once, store it next to the series, and let it gate which changes you look at.

Interpretation of an abnormal result, especially a persistent one, belongs with a clinician who can put it next to your history and examination. The value of doing your own arithmetic is knowing which results are worth that conversation.

What a standard panel cannot see

Standard chemistry indexes organ function at a fairly coarse level, and the informative frontier is in protein and nucleic-acid markers that most panels do not carry. Neurofilament light chain, an axonal cytoskeletal protein measurable in blood by single-molecule immunoassay, tracks neuroaxonal injury across a wide range of neurological conditions and is now a well-characterized quantitative marker rather than a research curiosity.² In cognitive assessment, plasma markers including phosphorylated tau species can distinguish Alzheimer-related from non-Alzheimer-related cognitive deficits, which no CMP component approaches.³ Circulating tumor DNA and protein markers in peripheral blood are under active development for solid tumor detection and monitoring, with performance that varies sharply by tumor type and stage.⁴ Transcriptomic panels extend further still: blood mRNA signatures have been trained to classify acute and chronic insufficient sleep in humans with machine learning, a physiological state no routine chemistry captures.⁵ None of these are screening tools for an asymptomatic person today, and ordering them without a clinical question produces numbers you cannot act on.

Getting your results as data

Ask the lab for results in a structured export. Do not settle for a PDF. In the United States, information-blocking rules under the 21st Century Cures Act mean most large labs expose a patient API, and a FHIR Observation resource per analyte is the cleanest form: it carries a LOINC code, a value, a unit in UCUM notation, the reference range, and the collection timestamp. LOINC codes matter because analyte names are ambiguous across labs while codes are not. Creatinine in serum is 2160-0, HbA1c is 4548-4, and apoB is 1884-6. Key on the code, store the raw text name as a secondary field, and you can merge panels drawn years apart at different labs.

Two data hygiene points save real pain. First, record units explicitly and convert once at ingest. Creatinine in mg/dL versus µmol/L differs by a factor of 88.4, and glucose in mg/dL versus mmol/L by 18.0. A silent unit switch between labs looks exactly like a step change in your physiology. Second, store the method and the performing lab alongside every value. Ferritin, TSH, and testosterone assays are not interchangeable across platforms. A platform change is the most common cause of an apparent break in a long series.

A minimal schema that has served us well is one row per observation with the following fields.

• loinc
• analyte
• value
• unit
• collected_at
• lab
• method
• ref_low
• ref_high
• flags (including the hemolysis index if reported)

Load it into a dataframe, join a static table of CVi values, and compute RCV per analyte. Then plot each series with the RCV band around the previous point rather than the population reference interval. That single change in visualization removes most false alarms.

If phlebotomy access is the constraint, dried blood spots are a viable collection route for a subset of analytes. They have a long track record in population research for markers including C-reactive protein and HbA1c. The cost is narrower analyte coverage and more variable precision than venous serum.⁶

Questions people also ask

Is it worth getting a full blood panel? Once, yes, as a structural baseline. After that, test on a schedule tied to what you are tracking. A single panel is a point estimate against a population range. The information compounds when you have four or five, because you can then compare each value to your own set point rather than to a distribution of strangers.

Will a comprehensive metabolic panel show A1c? No. The CMP includes a fasting glucose, which is a single-moment measurement, and HbA1c must be ordered separately. They answer different questions, and you want both.

What diseases can a comprehensive metabolic panel detect? A CMP is not a disease detector. The panel flags patterns consistent with reduced kidney filtration and with hepatocellular or biliary injury. It also flags electrolyte and acid-base disturbance, hypercalcemia, and hyperglycemia. Any of those patterns has several possible causes. Sorting them requires history, examination, and follow-up testing with a clinician.

Should I be worried if my comprehensive metabolic panel is abnormal? A single mildly out-of-range value is expected on a multi-analyte panel. It is usually resolved by repeating the draw under standardized conditions, in the same posture and at the same time of day. Avoid recent hard exercise or alcohol before that repeat draw. Persistent abnormalities, or any value far outside the range, should go to a clinician promptly.

What not to do before the draw? Do not clench your fist and do not let the tourniquet sit for more than a minute. Avoid iron or B12 supplements within 24 hours and alcohol within 48 hours. Do not train hard the day before. Do stay normally hydrated, since dehydration concentrates everything protein-bound.

Footnotes

1. Olle Melander. Novel and Conventional Biomarkers for Prediction of Incident Cardiovascular Events in the Community. JAMA, 2009. https://doi.org/10.1001/jama.2009.943 ↩

2. Lorenzo Gaetani, Kaj Blennow, Paolo Calabresi, et al. Neurofilament light chain as a biomarker in neurological disorders. Journal of Neurology, Neurosurgery & Psychiatry, 2019. https://doi.org/10.1136/jnnp-2018-320106 ↩

3. Kelsey R. Sewell, Lauren E. Oberlin, Thomas K. Karikari, et al. Blood biomarkers differentiate AD‐related versus non‐AD‐related cognitive deficits. Alzheimer’s & Dementia, 2025. https://doi.org/10.1002/alz.14619 ↩

4. Xinru Tu, Mengyan Tu, Junfen Xu. Clinical Application of Peripheral Blood Biomarkers for Solid Tumors. MedComm, 2026. https://doi.org/10.1002/mco2.70654 ↩

5. Emma E Laing, Carla S Möller-Levet, Derk-Jan Dijk, et al. Identifying and validating blood mRNA biomarkers for acute and chronic insufficient sleep in humans: a machine learning approach. Sleep, 2018. https://doi.org/10.1093/sleep/zsy186 ↩

6. Thomas W. McDade, Sharon Williams, J. Josh Snodgrass. What a drop can do: Dried blood spots as a minimally invasive method for integrating biomarkers into population-based research. Demography, 2007. https://doi.org/10.1353/dem.2007.0038 ↩