A laboratory reference interval is the band that the middle 95 percent of a chosen healthy population falls inside. It describes that population. It is not a verdict on you.
A clinical decision limit is a different object. It is a threshold a guideline committee sets for action, and it shifts with risk category and with the country whose guideline you are reading. Most of the confusion in a forty page screening report comes from treating those two numbers as the same kind of thing.
What is a reference interval, and why is it not a diagnosis?
It is a statement about a group of apparently healthy people, not about disease.
Reference intervals are derived from a reference distribution, usually the central 95 percent of it. The lower limit is the 2.5th percentile and the upper limit is the 97.5th percentile.
Two consequences follow from the arithmetic alone. Five percent of that healthy reference population sits outside the interval by construction. On a broad panel, a flagged value in a healthy person is expected rather than surprising.
The Clinical and Laboratory Standards Institute guideline that laboratories follow asks for at least 120 reference subjects before a laboratory can calculate confidence intervals around its own limits.
The review by Ozarda in Biochemia Medica sets out that requirement and separates the two concepts in one line. Reference intervals describe a specific population. Clinical decision limits are thresholds derived from receiver operating characteristic curves.
| Question | Reference interval | Clinical decision limit |
|---|
| What it answers | Where do most healthy people fall? | At what value does a guideline recommend acting? |
| Who sets it | The laboratory, from its own reference population and its own analyzer | A guideline panel, from outcome studies |
| What moves it | Population, age, sex, method, calibration | Risk category, new trial evidence, national guideline |
| Printed on the report | Yes, next to the result | Usually not |
The interval on the page belongs to the laboratory that ran the sample. Whether any single value matters for you is a question for you and your physician, not for the page.
Why do two laboratories print different ranges for the same test?
Because the reference population and the measurement method are both part of the definition.
A global multicenter study coordinated through the International Federation of Clinical Chemistry collected reference values for 50 serum analytes from 13,396 healthy individuals across 12 countries. It found sex related and age related patterns that were largely consistent worldwide. It also found that the relationship between body mass index and analyte level differed greatly between countries for some analytes.
Ozarda's review reports the same split by analyte.
| Analyte | Is a common worldwide interval feasible, per Ozarda's review |
|---|
| AST | Looked feasible |
| ALT | Looked feasible |
| Gamma glutamyl transferase | Population differences too large, so not applicable |
A 2026 study in BMC Cancer shows how far this can go on a single marker. The authors worked from 9,871 apparently healthy Han adults in China and a Beckman Coulter DxI 800 analyzer. They derived upper reference limits for serum CA-125 by age band.
| Women aged | Upper reference limit, U/mL | Widely quoted threshold, U/mL |
|---|
| 18 to 30 | 33.1 | 35 |
| 31 to 70 | 28.1 | 35 |
| 71 to 90 | 21.8 | 35 |
All three sit below the 35 U/mL threshold that is widely quoted, and the gap is widest in the oldest group.
So a number that reads as abnormal against one laboratory's limit can read as normal against another laboratory's limit, with neither laboratory having done anything wrong.
Is an LDL cholesterol number a laboratory range or a treatment target?
It is a treatment target, which is why it differs between countries without either country being in error.
The 2018 cholesterol guideline from the American Heart Association, the American College of Cardiology and ten other societies uses LDL cholesterol values as decision points rather than as normal ranges. Three of its thresholds show the pattern.
| Situation in the 2018 guideline | LDL-C, mg/dL | LDL-C, mmol/L | What the guideline says |
|---|
| Severe primary hypercholesterolemia | 190 or above | 4.9 | The entry point for the category |
| That group, already on high intensity statin therapy | 100 or above | 2.6 | Adding ezetimibe is reasonable |
| Very high risk atherosclerotic cardiovascular disease | 70 | 1.8 | Consider adding nonstatin therapy to a statin |
Taiwan has its own guideline. The 2022 Taiwan lipid guidelines for primary prevention were developed by the Taiwan Society of Lipids and Atherosclerosis with seven other societies, and the recommended LDL cholesterol target there is based on risk stratification as well.
Read that way, a lipid number carried between two countries is not one number with two right answers. It is one measurement being compared against two sets of thresholds written for two populations and two risk models. Which threshold applies to you is a clinical judgment your own physician makes.
What happens when the units are different?
The number changes and the meaning does not.
The 2018 cholesterol guideline prints both unit systems side by side for exactly this reason, which is why 190 mg/dL and 4.9 mmol/L name the same threshold in the table above.
Glycated hemoglobin is the sharper example, because the two systems produce numbers that look nothing alike. The National Glycohemoglobin Standardization Program publishes the master equation linking its own percentage scale to the International Federation of Clinical Chemistry scale in mmol/mol. The relationship is NGSP equals 0.09148 multiplied by IFCC, plus 2.152.
| NGSP, percent | IFCC, mmol/mol |
|---|
| 6.0 | 42 |
| 7.0 | 53 |
| 8.0 | 64 |
| 10.0 | 86 |
The NGSP notes that IFCC results run consistently lower than NGSP results across the range, and that there is worldwide consensus that the value should be reported in both. If your report gives one and your physician reads the other, the first job is conversion, not interpretation.
Why can the same tumor marker sample read differently on a different machine?
Because tumor marker assays are not fully harmonized across manufacturers.
A 2024 feasibility study in Clinical Chemistry pulled results from six external quality assessment programs in Germany, South Korea, China, the United Kingdom, the Netherlands and Australia. It covered between 19,040 and 25,398 individual results per marker on the Abbott Alinity, Beckman Coulter DxI, Roche Cobas and Siemens Atellica platforms.
The findings split the markers.
| Tumor marker | Harmonization status in that study |
|---|
| Alpha fetoprotein | Within the optimum bias criterion |
| Prostate specific antigen | Within the desirable criterion |
| Carcinoembryonic antigen | Minimum criterion overall, exceeded below 8 micrograms per liter |
| CA125, CA15-3 and CA19-9 | Outside the minimum bias criterion |
For those last three the study reported systematic differences between measurement procedures.
Calibration standards move the numbers too. When the same Beckman Coulter total PSA assay was recalibrated from the traditional Hybritech standard to the World Health Organization reference material, results across 1,098 patients in four centers came out 20 percent to 25 percent lower.
A separate study in Clinical Chemistry measured a 20.3 percent decrease and modelled the effect on a cohort of 5,865 men. Keeping the old biopsy thresholds after the change produced roughly a 20 percent fall in biopsies and a comparable fall in cancers detected.
The practical consequence is the one those authors drew. A report should say which manufacturer's assay and which calibration standard produced the number, and serial values should be compared only against values produced the same way.
Do professional bodies recommend tumor markers for screening healthy people?
Broadly, no, and that is worth saying plainly even though marker panels are widely sold.
The National Cancer Institute states that studies of circulating tumor markers for screening "have generally found that these markers do not work well for screening", because they are neither sensitive enough nor specific enough. It also states that "having an elevated level of a tumor marker does not mean that someone has cancer", since noncancerous conditions raise them.
The United States Preventive Services Task Force went further on one marker. In its 2018 statement it recommends against screening for ovarian cancer in asymptomatic women, a grade D recommendation, after concluding that screening with CA-125, transvaginal ultrasound, or both did not reduce ovarian cancer mortality in the major trials it reviewed.
None of that makes a marker meaningless in a person under investigation or under surveillance. It does mean that a single raised marker on a screening panel is a starting point for a conversation with a physician and nothing more.
If you are weighing whether to add a marker panel at all, our write up of tumor marker testing in Taiwan sets out what the panels contain.
What is a United States laboratory report required to contain?
Knowing this is the fastest way to see what a foreign report is missing.
Under the federal CLIA regulation at 42 CFR 493.1291, a test report must carry the patient's name and identification number or a unique patient identifier, "the name and address of the laboratory location where the test was performed", the test report date, the test performed, the specimen source when appropriate, and "the test result and, if applicable, the units of measurement or interpretation, or both".
Reports must also flag specimens that did not meet the laboratory's acceptability criteria.
Reference intervals are handled separately in the same section. The regulation requires that "pertinent reference intervals or normal values, as determined by the laboratory performing the tests" be available to the person who ordered the test and to those responsible for using the results.
| Item on the report | Why it matters when you take the report home |
|---|
| Name and address of the performing laboratory | Identifies whose reference interval applies and who can be asked about method |
| Units of measurement | Determines whether the number needs converting before comparison |
| Reference interval used | Lets a second reader see the limit the flag was set against |
| Assay or platform, where stated | Explains why a repeat value on another machine can differ |
| Date of collection and date of report | Sets the baseline for any future comparison |
What should you take to the appointment at home?
The Centers for Disease Control and Prevention advises travelers who receive care abroad to "request copies of their overseas medical records in English and to provide this information to any healthcare professionals they see subsequently for follow-up".
Working backwards from the CLIA list turns that into a short packing list.
- The full report rather than a summary, including the reference interval printed beside each result.
- The units for every numeric result, and the assay or platform where the report names it.
- The name and address of the laboratory or facility that ran each test, with contact details.
- An English copy obtained before you leave rather than translated by you afterwards.
- Any imaging in its original format rather than as a printed picture, alongside the written report.
Four questions are worth writing down before the appointment.
- Which values are being read against a limit that assumes a different population?
- Which need converting into the units your physician uses?
- Which would change management if they were confirmed?
- Which are worth repeating on a local assay before anyone acts?
If a scan rather than a blood test is the thing in question, our piece on incidental findings covers what usually happens next.
What this costs in Taipei
New Dawn Health arranges screening at eight core tiers.
| Screening tier | Price |
|---|
| Convenient | $299 |
| Essential | $699 |
| Standard | $1,199 |
| Light | $1,399 |
| Complete | $1,699 |
| Premium | $1,699 |
| Plus | $3,099 |
| Advanced | $3,499 |
What each tier contains is listed on our packages page.
The digital report is delivered 2 to 3 weeks after screening. An English physician debrief is available on request. Screening dates are arranged case by case, so confirm the slot before booking a flight, and ask if your dates are tight.
The bottom line
Sort the report into three piles before you ask anyone anything. Values inside their interval. Values outside an interval that belongs to one specific laboratory and one specific population. Values being compared against a guideline threshold rather than a range.
The second pile is where units, assay and reference population do most of the work. The third pile is where a national guideline and your own risk category decide what the number means. What to do about any of it is a decision for you and your own physician.
Sources
Blog
