Common laboratory reference values

Lab reports do not contain a single "normal" number for any test. They contain a reference interval that describes where the middle 95 per cent of a healthy reference population lies. Around one result in twenty from a truly healthy person falls outside that interval, so a lone value just outside range is not a diagnosis. The most useful number on the report is the interval printed next to the result for that specific laboratory, because ranges differ between methods, instruments, ages, sexes and physiological states. Always use the local interval, interpret the result in clinical context, and remember that three analytes only (cholesterol, creatinine and glycated haemoglobin) have been internationally harmonised: for everything else method-to-method variation can exceed 30 per cent. ^{[1][2] [7][9][12]}

By the end of this page you should be able to:

1. Define a reference interval and explain why 5 per cent of healthy people land outside it.
2. Use the tables below to look up conventional and SI units for common adult tests and convert between the two systems.
3. Say which factors alter a range (laboratory, age, sex, pregnancy, fasting, posture, time of day).
4. Recognise a critical (panic) value and list the common interpretation traps.
5. Apply a simple algorithm when a routine result is abnormal.

A reference interval is the range of values seen in a defined reference population on a defined test system. It is conventionally the central 95 per cent of results from 120 or more apparently healthy people drawn from that population, with the lowest 2.5 per cent and highest 2.5 per cent excluded. ^[1][9][12] It is therefore a statistical description of health, not a boundary between health and disease. Values in the excluded tails still occur in healthy people, while many patients with disease have results inside the interval. ^[7]

By definition, 5 per cent of healthy individuals will have at least one result outside the reference interval even when nothing is wrong, and in a panel of 20 tests about 64 per cent of healthy people will have one abnormal flag. ^[1] So an isolated result slightly outside range, especially in an asymptomatic person, is usually noise. Repeat it or interpret it in context before acting. ^[1]

"Normal" is also a misleading word. Healthy people can have results above or below the printed range, and people with disease can have results inside it. ^[7] The more precise term is "reference interval for a defined population on this test system".

Three kinds of variation dominate.

Method and analyser. Only cholesterol, creatinine and glycated haemoglobin (HbA1c) have been internationally harmonised, and even those show roughly 35 per cent spread between the lower and upper acceptable limits in proficiency schemes. ^[1] Troponin I still varies more than 4-fold. ^[1] For everything else, results and intervals depend on the method, the reagent and the instrument used, so two laboratories reporting the same analyte may print different limits.

Population. Age, sex, ethnicity and physiological state all move the reference population. Haemoglobin illustrates this well: it is 13.0-18.0 g/dL in adult men and 12.0-16.0 g/dL in adult women in one published adult system. ^[5] Alkaline phosphatase runs higher in growing children and higher after meals. ^[1] Advanced age, non-binary and transgender populations all have particular considerations. ^{[1] [13]}

Pre-analytical. Sample type (serum versus plasma), posture, time of day, fasting state, recent exercise and haemolysis change values. Glucose from a portable glucometer generally reads lower than plasma glucose from the main laboratory, and blood gas analysers in the emergency department often read haemoglobin 0.5-1.0 g/dL (5-10 g/L) higher than the central laboratory. ^[1] Always compare like with like and use the interval of the laboratory that ran the test. ^[2][9][12]

Conventional (mass/volume) units dominate US laboratory reports; SI (amount-of-substance) units dominate many other countries. The Mayo Clinic conversion table is a practical bridge. ^[4]

• Conventional unit x conversion factor = SI unit.
• SI unit divided by conversion factor = conventional unit.
• mEq/L is numerically identical to mmol/L for monovalent ions: sodium, potassium, chloride, bicarbonate. ^[4]
• For blood gases, 1 mmHg = 0.1333 kPa. ^[4]
• Glucose: mg/dL to mmol/L uses 0.05551. ^[4]
• Creatinine: mg/dL to mmol/L uses 0.0884.
• Calcium total: mg/dL to mmol/L uses 0.25. ^[4]
• Triglycerides: mg/dL to mmol/L uses 0.0113. ^[4]
• Total T4: mcg/dL to nmol/L uses 12.87; total T3: ng/dL to nmol/L uses 0.0154.
• Testosterone: ng/dL to nmol/L uses 0.0347; cortisol: mcg/dL to nmol/L uses 27.6.
• Insulin: mcIU/mL to pmol/L uses 6.95.
• Osmolality: mOsm/kg = mmol/kg.

Care is needed with glucose specimens. Serum and plasma values are not identical to whole-blood glucose, and every method has its own calibration. ^[1] In practice, always read the unit printed on the result and use the factor supplied with that analyser.

Table 1 is derived from an extended comprehensive metabolic panel catalog and is representative of an adult reference system; local intervals must take priority. ^[3]

Reference: ^[3]

The creatinine numbers use the standard factor 0.0884 mg/dL to mmol/L: multiply mg/dL by 0.0884 to obtain mmol/L and by 88.4 to obtain micromol/L. ^[3][4] Creatinine has been harmonised, but method spread of about 35 per cent still exists between laboratories. ^[1]

In pregnancy, both urea and creatinine fall by about 25 per cent in the first and second trimester and then trend back toward baseline. ^[10] No eGFR equation has been validated in pregnancy, so do not calculate eGFR or use formula thresholds there. ^[10]

The glucose interval shown is the catalog value for 1 year and over, which is deliberately wide because it includes non-fasting samples; a true fasting reference is much tighter and varies by laboratory. ^[3] The diabetes decision limits are given to show the difference between a biological reference interval and a diagnostic cut-off. The prediabetes and diabetes thresholds quoted (fasting glucose 100-125 mg/dL or 5.6-6.9 mmol/L, HbA1c 5.7-6.4 per cent or 39-47 mmol/mol) are the ADA 2026 criteria as reflected in the lattice reference set. ^[2]

Pregnancy lowers both adjusted and total calcium a little. Ionised calcium is unchanged. Magnesium falls about 30 per cent by the third trimester and phosphate falls slightly. ^[10] Bicarbonate runs about 25 per cent lower in pregnancy. ^[10]

The liver panel below uses adult intervals from the comprehensive metabolic panel catalog. ^[3]

Two facts separate lab values from disease. First, a "true healthy" ALT may be lower than many printed ranges: prospectively studied populations without identifiable liver risk factors have a healthy ALT of 29-33 IU/L in males and 19-25 IU/L in females, so values above that warrant assessment even when they sit inside the laboratory interval. ^[18] Second, ALP measured after a meal is higher than a fasting or basal value, and alkaline phosphatase is made by bone as well as liver, so an isolated rise in ALP should be confirmed hepatic with GGT before investigating the bile ducts. ^[1][3]

In pregnancy, ALT, AST, GGT and bilirubin fall slightly, while ALP rises because of the placental isoenzyme. ^[10] A single pregnancy-ready reference interval for any of these does not exist; the safest message is to interpret against the laboratory's own pregnancy-adjusted interval where provided. ^[10]

MedlinePlus publishes thresholds produced with the American Heart Association rather than a pure reference interval; lipids have been reclassified over time as outcome evidence accumulated. ^[1][15][16]

Conversion uses the Mayo Clinic table (0.0113 for triglycerides and 12.87 for total cholesterol conversion via the standard factors in the cited references). ^[4] In pregnancy, total and LDL cholesterol double and triglycerides triple by late pregnancy, so a pregnant woman's lipid panel is uninterpretable against non-pregnant thresholds. ^[10]

Thyroid hormone intervals are among the least portable results: TSH varies about 35 per cent between laboratories even when the same method is used, and no fixed pregnancy thresholds exist. ^[1][10][17]

TSH values vary during the day and are best measured in the morning. Some laboratories use a higher upper limit for older people (even as high as 7 microunits/mL), and experts disagree on the exact upper limit of normal in thyroid diagnosis. ^[17] In pregnancy, thyroxine-binding globulin rises 2-3-fold and total T4 rises accordingly; institutions are urged to compute their own trimester- and method-specific intervals rather than apply fixed cut-offs. ^[10]

The two adult CBC reference systems below differ because they come from different populations and analysers: both are authentic, published adult systems. ^[5][2] The message is that the laboratory's own system is the right one to use.

Both sets overlap but are not identical, which is exactly the point about portability. ^[5][2]

Pregnancy dilutes the blood: plasma volume rises more than red cell mass, giving a physiological anaemia. A haemoglobin under 10.5 g/dL (105 g/L) or a haematocrit under 32 per cent is considered anaemia in the second trimester; the first and third trimester cut-offs are 11.0 g/dL and 33 per cent. ^[11] Ferritin falls to about 50 per cent of its pre-pregnancy nadir by the third trimester. ^[10] Always use a pregnancy-specific interval.

PT and INR from MedlinePlus; aPTT interval from MedlinePlus. ^[13][14] Warfarin targets can be wider for mechanical valves and certain thrombophilias, so use the indication-specific range rather than 2.0-3.0 alone. Prothrombin time prolongs early when the liver stops making clotting factors, which makes PT and INR a more sensitive early marker of hepatic synthetic failure than albumin. ^[6]

Ferritin thresholds from the lattice set; total iron, TIBC and haematinic values from the same set. ^[2]

Values from MedlinePlus blood gas article. ^[6]

Urine values from MedlinePlus urinalysis material. ^[8] Cerebrospinal fluid glucose is commonly greater than two-thirds of concurrent blood glucose, a useful cross-check when reading paired samples. ^[8] Urine specific gravity differs between sources (1.005-1.030 versus 1.002-1.030), another illustration of method-dependent intervals. ^[2][8]

A critical value (panic value) is a result so far from the reference interval that it is potentially life-threatening and requires immediate notification of the responsible clinician. Laboratories define their own; common examples include: ^[9]

• Sodium under 120 or over 160 mmol/L.
• Potassium under 2.5 or over 6.5 mmol/L.
• Glucose very low (about 40 mg/dL or 2.2 mmol/L) or very high (about 400-500 mg/dL or 22-28 mmol/L).
• Haemoglobin under 7.0 g/dL (70 g/L).
• Platelets extremely low.
• INR markedly elevated in a patient not on anticoagulation.
• Arterial pH under 7.20 or PaCO2 over 60 mmHg (8 kPa).
• Troponin above the assay-specific 99th percentile.

Precise numeric cut-offs are laboratory-specific, so learn your own institution's list; every accredited laboratory publishes it with its reference range. ^[9]

1. One abnormal flag is expected. As above, 64 per cent in a 20-test panel of healthy people. ^[1]
2. Hourly variation is built into some tests. Cortisol, TSH, testosterone and growth hormone have diurnal patterns; sample at the wrong time and the result misleads.
3. Enzymes change with activity and meals. Creatine kinase rises with exertion; alkaline phosphatase rises after eating. ^[1]
4. Age-related drift. Many analytes (PSA, enzymes, ESR, TSH) rise with age; children are not small adults. ^[1]
5. Sex modifies HDL, creatinine, ferritin and haemoglobin. Use the sex-appropriate interval. ^[1]
6. Method differences between emergency and central labs. Differences of 0.5-1.0 g/dL (5-10 g/L) for haemoglobin are real. ^[1]
7. Ethnicity and eGFR. Recent equations removed the ethnicity factor; some physiological differences remain. ^[1]
8. Pregnancy changes everything. Creatinine, urea, osmolality and bicarbonate fall; ALP and lipids climb two- to three-fold; thyroid-binding globulin rises. Use pregnancy-specific or trimester-specific intervals. ^[10]
9. True healthy cut-offs can be tighter than the printed interval. ALT, BMI, blood pressure and cholesterol have all had their "normal" thresholds lowered as evidence accumulated. ^[1]
10. Reference intervals are descriptions, not prescriptions. In some analyses a Non-HDL cholesterol of 200 mg/dL or a BMI of 30 has been linked with lower mortality than lower values. ^[1]
11. 99th percentile troponin is a one-sided cut-off rather than a two-sided range. ^[1]

A simple memory aid: an abnormal flag means "investigate", not "disease".

Children and adolescents. Children are not small adults. Alkaline phosphatase runs higher while growth plates are open, haemoglobin and haematocrit change through childhood, and renal function matures. Published pediatric intervals should be used right up to the age limit printed on the report rather than an adult interval. ^[1][9]

Older adults. Many analytes drift upward with age (PSA, creatinine, ESR and sometimes TSH), so an older patient whose result sits at the upper half of the adult interval may still be within his or her age-adjusted expectation. Some laboratories use a higher upper limit for TSH in older people, as high as 7 microunits/mL. ^[1][17]

Non-binary and transgender people. Reference ranges for non-binary and transgender patients are still being worked out; where published intervals are unavailable, the appropriate anchor is the sex-typical interval for the physiology predominating at the time, interpreted with the clinician. ^[1]

Altitude and geography. PaO2 falls at 3000 feet (900 metres) and above, and haemoglobin can be higher because of chronic hypoxia. Reference populations differ by ethnicity too: some physiological values (vitamin D binding protein, vitamin D levels) differ without indicating disease. ^[1][6]

Ethnicity and race. Race-based modifiers have been removed from eGFR calculations because race is a social construct, and physiological differences should be interpreted as biology, not labels. ^[1]

Pre-analytical conditions. Fasting 9-12 hours for lipids and glucose tolerance tests, morning sampling for cortisol, TSH, testosterone and growth hormone, 20-30 minutes seated rest before a full blood count and lipid sample, avoidance of recent heavy exercise before creatine kinase and lactate dehydrogenase, and consistent use of the same laboratory for serial testing all reduce apparent "abnormality" that is really variation. ^[1][13]

When a value changes from a patient's own baseline, the size of that change matters more than whether it crosses the printed interval edge. A creatinine that rises from 0.7 to 1.1 mg/dL (62 to 97 micromol/L) in an older man is clinically significant even though 1.1 mg/dL is inside the printed range. Repeat testing from the same laboratory, with the same method, is the only way to make those comparisons safely, because method-to-method variation can exceed 30 per cent even for harmonised analytes. ^[1] Delta values, where the laboratory reports the change from the prior result, are more useful for monitoring than repeated comparison against a population interval.

• Use the interval printed on the local report, not a textbook or internet range; if none is printed, contact the laboratory.
• Keep serial testing in the same laboratory, with the same reagent lot where possible, to reduce between-lab variation.
• If an unexpected abnormal flag appears in an asymptomatic patient, repeat it before acting; one in twenty healthy people flag positive on a single test and most flags are noise.
• Use physiological knowledge as a check: in pregnancy, a creatinine of 0.9 mg/dL may already be high; at altitude, a lower PaO2 is expected.
• For suspected hepatic synthetic failure, use PT/INR rather than albumin in the first day: prothrombin time falls when clotting factor synthesis drops, so INR moves early, whereas albumin changes slowly.
• Where only a glucose strip is available, remember that whole-blood glucose generally reads lower than plasma glucose; confirm with the laboratory method when possible. ^[1]

• A reference interval is the central 95 per cent of values in a defined healthy population on a defined test system. ^[1]
• About 5 per cent of healthy people fall outside it; about 64 per cent have one flagged abnormality in a 20-test panel. ^[1]
• Only cholesterol, creatinine and HbA1c are internationally harmonised. ^[1]
• Use the interval printed on the report of the laboratory that ran the test. ^[2][9][12]
• Conventional and SI units convert by fixed factors; mEq/L equals mmol/L for monovalent ions. ^[4]
• Age, sex, fasting, posture, time of day, exercise, altitude and pregnancy all shift values. ^[1][10]
• Pregnancy lowers creatinine, urea, osmolality, bicarbonate and albumin and raises ALP and lipids two- to three-fold. ^[10][11]
• Critical values require immediate action; each laboratory publishes its own list. ^[9]
• An ALT above 29-33 IU/L in men or 19-25 IU/L in women warrants assessment even when it sits inside the laboratory interval. ^[18]
• TSH (0.4-4.8 mIU/L) and total T4 are strongly assay- and pregnancy-dependent, so method-specific limits are needed. ^[10][17]
• Where sources or guidelines differ, published thresholds always refer to their own methods and populations, and local intervals take priority. ^[1][10]