A result can look worse after a lab, country, app, or reporting format changes units. The clinical question is whether the biology changed — not whether the number looks bigger.
- mg/dL vs mmol/L can make the same glucose result appear as 90 mg/dL or 5.0 mmol/L; the body has not changed.
- LDL cholesterol converts by dividing mg/dL by 38.67, so 116 mg/dL is about 3.0 mmol/L.
- Glucose converts by dividing mg/dL by 18.02, so 126 mg/dL is about 7.0 mmol/L and meets a diabetes diagnostic threshold when confirmed.
- IU/L and U/L are usually equivalent for enzyme reporting, but ALT or AST values can still differ because assay methods and reference ranges vary.
- CBC percentages can look high even when absolute counts are normal; a lymphocyte percentage of 48% may be harmless if the absolute lymphocyte count is 2.4 x 10^9/L.
- Decimal places create false precision; creatinine 0.99 mg/dL and 1.0 mg/dL are usually the same clinical result.
- Lab-specific ranges reflect local instruments, reagents, age groups, sex, pregnancy status, and statistical choices, not universal health boundaries.
- Kantesti AI checks units, reference intervals, trends, and biomarker patterns together so apparent changes are not mistaken for true deterioration.
Why lab values can look different when nothing meaningful changed
Lab values can look better or worse simply because the unit, decimal format, assay method, or reference range changed. A glucose of 90 mg/dL is the same biological value as about 5.0 mmol/L, and an LDL cholesterol of 116 mg/dL is about 3.0 mmol/L. As of May 2, 2026, this is one of the most common reasons patients misread international or online lab test results. I am Thomas Klein, MD, and in my clinical review work at Kantesti AI, unit changes are among the first things I check before calling a trend real.
The practical test is simple: ask whether the same analyte is being measured, in the same unit, by a similar method, and against a comparable reference interval. If any one of those 4 conditions changes, a 10% swing on the page may be paperwork rather than physiology.
I see this pattern when patients move between the UK, Canada, the US, the Gulf states, and Europe. One 44-year-old runner thought his cholesterol had “tripled” after moving labs; his report had switched from mmol/L to mg/dL, and his LDL was essentially unchanged after conversion.
Before reacting to a flag, compare the unit and the reference range printed beside the value. Our guide on how to read blood test results explains the same principle: a number without its unit is not a medical result.
How mg/dL and mmol/L change the look of blood test numbers
mg/dL reports mass per volume, while mmol/L reports molecule count per volume, so conversion factors differ for each marker. Glucose uses 18.02, cholesterol uses 38.67, triglycerides use 88.57, and calcium uses about 4.0.
A fasting glucose of 100 mg/dL equals about 5.6 mmol/L, because glucose’s molecular weight drives the conversion. According to the American Diabetes Association Standards of Care 2024, fasting plasma glucose of 126 mg/dL, or 7.0 mmol/L, meets a diabetes threshold when confirmed on repeat testing or supported by symptoms.
Cholesterol is where patients get caught most often. The 2018 AHA/ACC cholesterol guideline discusses LDL-C in mg/dL, while many European and Commonwealth labs report mmol/L; LDL-C 70 mg/dL is about 1.8 mmol/L, a common target in high-risk cardiovascular care (Grundy et al., 2019).
Triglycerides are not converted with the cholesterol factor. A triglyceride result of 150 mg/dL is about 1.7 mmol/L, and using the LDL conversion by mistake can make the value look wildly wrong.
Kantesti’s blood test biomarkers guide stores biomarker-specific unit logic because a single “divide by 18” rule is unsafe. For cholesterol-specific interpretation, our lipid panel guide shows how LDL, HDL, and triglycerides should be read together.
Why IU/L, U/L and enzyme methods make liver results shift
IU/L and U/L usually mean the same thing for many enzyme tests, but the reported number can still change when the lab uses a different assay temperature, reagent, calibration, or reference range. ALT 42 U/L at one lab and 48 IU/L at another may not represent new liver injury.
ALT, AST, ALP, GGT, CK, and LDH are activity measurements, not molecule counts. A result of AST 89 U/L in a 52-year-old marathon runner after a hard race may reflect skeletal muscle stress, while the same AST with bilirubin 3.0 mg/dL and INR 1.6 feels very different clinically.
Some European labs use lower ALT upper limits, often around 35 U/L for men and 25 U/L for women, while other reports still show upper limits near 40–56 U/L. That range choice can flip a flag from normal to high without any biological movement.
The reason we worry about ALT plus bilirubin is pattern, not the isolated unit. For liver patterns, compare enzyme families using our liver function test guide before assuming one mildly high enzyme means liver disease.
When I review serial liver panels, I treat a change from ALT 41 to 45 U/L as noise unless symptoms, bilirubin, ALP, GGT, medications, alcohol exposure, or imaging point in the same direction. For a deeper ALT-specific read, see our ALT blood test guide.
Why percentages can look alarming when absolute counts are normal
Percentages show the share of a cell type, while absolute counts show how many cells are actually present. In a CBC, a lymphocyte percentage of 48% can be normal if the absolute lymphocyte count is within about 1.0–4.0 x 10^9/L.
A percentage always depends on the denominator. If neutrophils temporarily fall after a viral illness, lymphocyte percentage may rise even though the absolute lymphocyte count has not increased.
I recently reviewed a CBC with lymphocytes 54% and a normal absolute lymphocyte count of 2.7 x 10^9/L. The patient had spent a weekend convinced this meant leukemia; the actual issue was a low-normal neutrophil count after a mild respiratory virus.
Absolute neutrophil count matters more than neutrophil percentage for infection risk. An ANC below 1.5 x 10^9/L is generally called neutropenia, and below 0.5 x 10^9/L is where infection risk becomes much more clinically serious.
Our CBC differential guide walks through these absolute-versus-percentage patterns. We also explain why high lymphocyte percentage often means less than patients fear when the absolute count is normal.
How decimal places and rounding create fake changes
Decimal places can make stable blood test numbers look newly abnormal, especially near cutoffs. Creatinine 0.99 mg/dL and 1.0 mg/dL are usually the same clinical result, even though one may display with 2 decimals and the other with 1.
Many analyzers measure more precision than the lab chooses to print. A potassium value measured as 4.44 mmol/L may be reported as 4.4 by one lab and 4.44 by another, which can make trend charts look more exact than the biology allows.
Thomas Klein, MD may sound fussy about rounding, but this matters. A TSH of 4.49 mIU/L may display as 4.5, and if the lab cutoff is 4.50, one report can be flagged while the other is not.
The medical phrase here is analytical variation, and it differs by test. HbA1c, creatinine, sodium, and CRP each have different acceptable imprecision; a change of 0.1% HbA1c means less than a change of 1.0% HbA1c in most adults.
For real-world thresholds, pair rounding with the bigger concept of biological variability. Our article on blood test variability gives practical examples, and borderline result interpretation helps when a value sits just over the line.
Why normal ranges differ between labs
Reference ranges differ because labs use different instruments, reagents, populations, and statistical methods. A normal range is usually the middle 95% of a selected reference group, so about 5% of healthy people can sit outside it.
A lab’s reference interval is not a perfect health boundary. It is often built from local data or manufacturer validation, then adjusted for age, sex, pregnancy, and sometimes method-specific calibration.
Kantesti reviews these ranges against clinical context rather than treating every flag as disease. Our medical validation standards describe how our clinical team handles method differences, and our Medical Advisory Board reviews safety-sensitive interpretation rules.
One quietly tricky example is ferritin. A lab may list adult female ferritin as 15–150 ng/mL, but a symptomatic patient with restless legs or hair shedding may feel better when ferritin is higher than the lower limit; clinicians disagree on the exact cutoff.
The flag is a clue, not a diagnosis. Our blood test normal range article explains why “normal” and “optimal for this person” are not always the same thing.
Why creatinine and eGFR change with units and formulas
Creatinine may be reported in mg/dL or µmol/L, and eGFR depends on the formula used as well as age and sex. Creatinine 1.0 mg/dL is about 88.4 µmol/L, but the same creatinine can produce different eGFR values in different people.
A creatinine result is not a kidney diagnosis by itself. A muscular 32-year-old man with creatinine 1.25 mg/dL may have normal filtration, while a frail 82-year-old woman with creatinine 1.0 mg/dL may have a reduced eGFR.
The CKD-EPI equation published by Levey et al. in Annals of Internal Medicine in 2009 improved eGFR estimation compared with older formulas, especially at higher GFR values. Many labs have since updated again to race-free equations, which can move eGFR by several mL/min/1.73 m² without any kidney change.
An eGFR below 60 mL/min/1.73 m² for at least 3 months supports chronic kidney disease when persistent or accompanied by kidney damage markers. A single eGFR of 58 after dehydration, creatine use, or heavy exercise is not the same as confirmed CKD.
For kidney trends, compare creatinine units, eGFR equation notes, urine albumin, and hydration status. Our eGFR by age guide and GFR versus eGFR explainer give the deeper kidney-specific context.
Why iron studies are especially easy to misread
Iron studies mix mass units, molar units, percentages, and binding capacity, so they are one of the easiest panels to misread. Ferritin may appear as ng/mL or µg/L, and those 2 units are numerically equivalent for ferritin.
Serum iron can be reported in µg/dL or µmol/L; to convert serum iron from µg/dL to µmol/L, multiply by about 0.179. A serum iron of 70 µg/dL is about 12.5 µmol/L, not 70 in a different system.
Transferrin saturation is a percentage, usually calculated as serum iron divided by TIBC times 100. A saturation below about 20% often supports iron deficiency when ferritin and symptoms fit, but inflammation can push ferritin upward and hide low available iron.
Ferritin is both an iron storage marker and an acute-phase reactant. I have seen ferritin 240 ng/mL in a patient with fatty liver and low iron availability; the “high ferritin” did not mean iron overload once saturation was 12%.
For this reason, Kantesti AI reads iron markers as a panel. The TIBC and iron saturation guide and our article on low ferritin with normal hemoglobin are useful when blood test numbers meaning depends on the pattern.
How hormone units make thyroid and testosterone results look different
Hormone results often switch between mIU/L, pmol/L, ng/dL, nmol/L, pg/mL, and µIU/mL. TSH in mIU/L and µIU/mL is usually numerically the same, but free T4 and testosterone need true conversion.
Free T4 is commonly reported as pmol/L outside the US and ng/dL in many US reports. A free T4 of 1.2 ng/dL is about 15.4 pmol/L, which is mid-range in many laboratories.
Total testosterone converts differently: 1 ng/dL is about 0.0347 nmol/L, so 500 ng/dL is about 17.3 nmol/L. Morning timing matters; testosterone can be 20–30% lower later in the day in younger men.
TSH interpretation is one of those areas where context matters more than one cutoff. A TSH of 4.8 mIU/L with normal free T4 may mean mild subclinical hypothyroidism in one person, recovery from illness in another, and a pregnancy-specific concern in someone else.
For thyroid patterns, read TSH beside free T4, free T3, antibodies, medication timing, and biotin exposure. Our thyroid panel guide and free testosterone guide show why unit conversion is only step 1.
Why vitamin D, B12 and folate numbers change by country
Vitamin units vary widely by country and assay, so the same nutrient status can look like a new deficiency or sudden improvement. Vitamin D 30 ng/mL equals 75 nmol/L, because 25-hydroxyvitamin D converts by multiplying ng/mL by 2.5.
Vitamin D cutoffs are genuinely debated. Many clinicians consider 20 ng/mL or 50 nmol/L adequate for bone health in much of the population, while others aim for 30 ng/mL or 75 nmol/L in higher-risk patients.
Vitamin B12 may be reported in pg/mL or pmol/L; to convert pg/mL to pmol/L, multiply by about 0.738. A B12 of 300 pg/mL is roughly 221 pmol/L, which can be borderline if symptoms, methylmalonic acid, or homocysteine suggest functional deficiency.
Folate can be reported as ng/mL or nmol/L, and red cell folate behaves differently from serum folate. A recent supplement can raise serum folate within days, while red cell folate reflects a longer window of about 120 days of red cell lifespan.
Kantesti’s neural network treats nutrients as time-sensitive markers, not isolated labels. For more context, see our vitamin D blood test guide and B12 normal range article.
Why INR, PT and anti-Xa use different reporting logic
Coagulation tests use seconds, ratios, percentages, and activity units, so comparing them like ordinary chemistry values is unsafe. INR is a standardized ratio, PT is measured in seconds, and anti-Xa is often reported in IU/mL.
INR was designed to make warfarin monitoring more comparable across thromboplastin reagents. For many warfarin indications, the target INR is 2.0–3.0, while some mechanical heart valves require a higher target chosen by the treating clinician.
PT seconds cannot be interpreted without knowing the reagent and lab. A PT of 14.5 seconds may be mildly high in one laboratory and near normal in another, which is why INR exists for vitamin K antagonist monitoring.
Anti-Xa is different again. A low molecular weight heparin anti-Xa level is often interpreted around 0.5–1.0 IU/mL for certain twice-daily therapeutic regimens, but timing after dose and drug type change the target.
Do not convert PT to INR yourself using an online calculator unless the lab provides the correct sensitivity data. Our PT/INR normal range guide and coagulation test explainer cover the safety details.
Why urine results use plus signs, ratios and mixed units
Urine test results may use mg/dL, mmol/L, mg/g, mg/mmol, plus signs, or descriptive categories. A urine protein “trace” result is not equivalent to a quantified albumin-creatinine ratio of 30 mg/g or 3 mg/mmol.
Dipstick urine protein is semi-quantitative, which means it estimates a range rather than a precise amount. Concentrated urine can read 1+ protein while the true daily albumin loss is not high, especially after exercise or fever.
Albumin-creatinine ratio helps correct for urine concentration. An ACR of 30 mg/g, or about 3 mg/mmol, is a common threshold for moderately increased albuminuria when persistent.
Urobilinogen is another example of confusing reporting. Some reports use EU/dL, others use mg/dL, and many labs simply flag increased or normal categories; hydration and liver-biliary context matter more than a lone borderline value.
For urine interpretation, compare dipstick, microscopy, ACR, kidney function, and symptoms. Our complete urinalysis guide explains the categories, while the renal function panel guide connects urine clues to blood chemistry.
How to compare trends when labs or countries change
Trend comparison is safest when you convert units first, then compare results against the same biological baseline and clinical context. A real trend usually persists across at least 2–3 measurements or matches symptoms, medications, imaging, or other biomarkers.
When patients bring me 5 years of reports from 3 countries, I do not start with flags. I build a same-unit timeline, mark fasting status, note medication changes, and separate method changes from biological shifts.
A true LDL improvement might be 4.1 to 2.6 mmol/L, which is about 158 to 101 mg/dL. A fake improvement might be 101 mg/dL to 2.6 mmol/L misread as a fall from 101 to 2.6 without conversion.
For inflammatory markers, the timing can matter more than the unit. CRP may fall from 48 mg/L to 6 mg/L after an infection resolves, while hs-CRP for cardiovascular risk is usually interpreted in a much lower range, often below 1, 1–3, and above 3 mg/L.
Use one trend chart whenever possible. Our blood test comparison guide and year-over-year history guide show how to avoid comparing apples, oranges, and decimal formatting.
How Kantesti AI checks units before interpreting results
Kantesti AI reads the unit, reference interval, biomarker name, date, and report layout before interpreting the result. Our AI-powered blood test interpretation checks more than 15,000 biomarkers and flags likely unit mismatches before trend conclusions are generated.
Our platform supports PDF and photo upload because real lab reports are messy. A scanned report may show creatinine in µmol/L, cholesterol in mmol/L, vitamin D in ng/mL, and CBC counts in 10^9/L on the same page.
Kantesti uses document parsing, biomarker dictionaries, population rules, and clinical logic to avoid obvious traps. The engine’s clinical benchmarking is described in our AI blood test benchmark, and a pre-registered validation update is available through our clinical validation DOI.
The output is still medical guidance, not a replacement for urgent care or your clinician. If potassium is 6.4 mmol/L, troponin is rising, hemoglobin is 6.8 g/dL, or INR is 7.0 with symptoms, the unit check should happen fast — but the next step may be emergency evaluation.
You can learn how the upload pipeline works in our blood test PDF upload guide. For image-based reports, our blood test photo scan article explains why lighting, crop quality, and missing reference ranges affect interpretation.
What to check before acting on a changed lab value
Before acting on changed lab values, confirm the unit, reference range, test method, date, fasting status, medication timing, and whether the absolute value or percentage matters. Most non-urgent results deserve a 5-minute verification step before anxiety, supplements, or dose changes.
My safety checklist has 7 steps: same biomarker, same unit, same lab or method, same reference population, same timing, same clinical state, and same medication context. If 2 or more changed, I treat the trend as provisional until verified.
Do not adjust thyroid medication, diabetes medication, anticoagulants, iron, potassium, or vitamin D solely because a number looks different in a new unit. A dose change based on a misread result can create real harm within days, especially with insulin, warfarin, potassium supplements, or levothyroxine.
Thomas Klein, MD advice here is deliberately plain: convert first, interpret second, act third. If the result is critically abnormal or symptoms are severe — chest pain, fainting, confusion, severe weakness, black stools, or shortness of breath — do not wait for an app or blog.
For non-urgent questions, upload your report to Try Free AI Blood Test Analysis and compare the interpretation with your clinician’s advice. You can also review our critical values guide to learn which lab test results need same-day attention.
Frequently Asked Questions
Why do my lab values look different on a new report?
Lab values often look different because the report changed units, decimal places, reference ranges, or laboratory method. A glucose of 90 mg/dL is about 5.0 mmol/L, so the smaller number is not an improvement or decline. Always compare the biomarker name, unit, reference interval, and date before judging a trend.
How do I convert mg/dL to mmol/L for blood test results?
mg/dL to mmol/L conversion depends on the specific molecule, so there is no single safe conversion factor. Glucose in mg/dL is divided by 18.02, LDL cholesterol is divided by 38.67, triglycerides are divided by 88.57, and calcium is divided by about 4.0. Using the wrong factor can make normal lab test results look dangerously abnormal.
Is IU/L the same as U/L on liver blood tests?
IU/L and U/L are usually treated as equivalent for many enzyme blood tests, including ALT, AST, ALP, GGT, CK, and LDH. The bigger source of difference is often the assay method, temperature, calibration, or lab-specific reference range. ALT 42 U/L at one lab and 48 IU/L at another may not represent a meaningful biological change.
Why is my lymphocyte percentage high but my count normal?
A lymphocyte percentage can be high when the absolute lymphocyte count is normal because percentages depend on the other white cell types. For adults, an absolute lymphocyte count around 1.0–4.0 x 10^9/L is commonly within range, even if the percentage is 45–55%. Doctors usually give more weight to the absolute count than the percentage alone.
Can rounding change whether a blood test is flagged high or low?
Yes, rounding can change the appearance of a flag near a cutoff. A TSH of 4.49 mIU/L may display as 4.5 mIU/L, and a creatinine of 0.99 mg/dL may display as 1.0 mg/dL. These are usually tiny analytic differences, not true medical changes, unless the pattern repeats or related markers also move.
Why do normal ranges differ between laboratories?
Normal ranges differ because laboratories use different instruments, reagents, calibration systems, populations, and statistical methods. Many reference intervals represent the middle 95% of a selected group, meaning about 5% of healthy people can fall outside the printed range. A value should be interpreted with age, sex, pregnancy status, symptoms, and prior baseline.
Can Kantesti AI detect unit mismatches in blood test reports?
Kantesti AI checks biomarker names, units, reference intervals, dates, and report formatting before generating an interpretation. The platform is designed to recognize common unit systems such as mg/dL, mmol/L, IU/L, %, ng/mL, µmol/L, and 10^9/L. It can help identify when a result looks changed because the unit changed, but urgent or severe results still require clinician review.
Get AI-Powered Blood Test Analysis Today
Join over 2 million users worldwide who trust Kantesti for instant, accurate lab test analysis. Upload your blood test results and receive comprehensive interpretation of 15,000+ biomarkers in seconds.
📚 Referenced Research Publications
Klein, T., Mitchell, S., & Weber, H. (2026). Urobilinogen in Urine Test: Complete Urinalysis Guide 2026. Kantesti AI Medical Research.
Klein, T., Mitchell, S., & Weber, H. (2026). Iron Studies Guide: TIBC, Iron Saturation & Binding Capacity. Kantesti AI Medical Research.
📖 External Medical References
American Diabetes Association Professional Practice Committee (2024). 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes—2024. Diabetes Care.
⚕️ Medical Disclaimer
This article is for educational purposes only and does not constitute medical advice. Always consult a qualified healthcare provider for diagnosis and treatment decisions.
E-E-A-T Trust Signals
Experience
Physician-led clinical review of lab interpretation workflows.
Expertise
Laboratory medicine focus on how biomarkers behave in clinical context.
Authoritativeness
Written by Dr. Thomas Klein with review by Dr. Sarah Mitchell and Prof. Dr. Hans Weber.
Trustworthiness
Evidence-based interpretation with clear follow-up pathways to reduce alarm.