Vitamin D3 for Life

Live Enlightened

Article

Analytical and Population-Level Noise in Vitamin D Test Interpretation: Full Study

Previous related pageA Physiology-First Framework for Vitamin DNext related pageAnalytical and Population-Level Noise in Vitamin D Test Interpretation: Study Summary
Study at a glance

Why can vitamin D results differ between tests and groups of people?

This page examines why vitamin D results differ between laboratory tests and between groups of people. The studies measured total 25-hydroxyvitamin D, the usual vitamin D blood marker. They included comparisons of tests on donated blood samples and analyses of population surveys.

Does a consistent vitamin D test result always mean an accurate result?

Researchers tested 50 blood samples with 13 laboratory tests. All 13 gave sufficiently consistent results when measurements were repeated, but only nine met the study’s accuracy limit. Some tests responded differently when samples contained more of the form made from vitamin D2. Repeating a similar result therefore did not guarantee that the result was accurate.

How did a shared measurement standard change the Irish vitamin D survey results?

In an Irish survey, bringing results onto a shared measurement standard changed the estimated percentage below 30 nmol/L. The estimate rose from 6.5% to 11.4%, close to the 11.2% found when samples were measured again using a standardised method. The change reflected measurement methods, rather than a new change in people’s health.

Can laboratory methods change which vitamin D category a result falls into?

Laboratory methods can affect whether a result falls above or below a chosen cut-off. Shared standards improve comparisons, but real differences between people and seasons remain. A range seen in one population is not automatically an ideal target for everyone. These findings do not mean all tests are unreliable or that every difference between results is a laboratory error.

Read the full article below ↓

Vitamin D test interpretation is affected by two problems that can look similar but arise for very different reasons. First, laboratories and assays can produce different 25-hydroxyvitamin D, or 25(OH)D, values from the same or comparable samples. Second, populations differ biologically and environmentally, so the distribution of measured values changes with season, geography, body composition, age, exposure patterns and other characteristics. A numerical result therefore sits inside both an analytical measurement system and a population context.

Analytical variability is not merely theoretical

Substantial assay variation has been documented repeatedly. The Vitamin D Standardization Program, or VDSP, was created in part because results for total serum 25(OH)D were not sufficiently comparable across methods and laboratories. Standardisation aims to align routine measurements with reference measurement procedures so that results represent the same measurand regardless of where and how they are measured.

A major VDSP intralaboratory study by Wise and colleagues tested 13 assays in one laboratory using 50 individual-donor serum samples. The assay set included 11 unique immunoassays and an LC-MS/MS method. Reference target values for 25(OH)D2 and 25(OH)D3 were assigned using reference measurement procedures.

This design was valuable because it separated variation caused by different commercial measurement systems from variation caused by different laboratories. Four replicate measurements were generated for each sample and assay.

Precision and bias are different problems

The VDSP performance criteria used in the study included a coefficient of variation no greater than 10 percent and mean bias within plus or minus 5 percent. All 13 evaluated assays achieved the overall precision criterion. Only nine met the mean bias criterion.

This distinction is important. A method can produce closely clustered repeated measurements and still be systematically displaced from the reference value. Good repeatability is therefore not proof of accuracy.

The mean total 25(OH)D concentration across the 50 donor samples varied substantially between assays. The NIST reference mean was 77.1 nmol/L, while assay means ranged from 69.0 to 93.3 nmol/L. That spread is large enough to alter interpretation when a value sits near a classification threshold.

Vitamin D2 recovery exposes an additional source of assay disagreement

The VDSP study intentionally included eight samples containing more than 30 nmol/L of 25(OH)D2. These samples acted as a stress test because a total 25(OH)D assay should appropriately measure both 25(OH)D2 and 25(OH)D3.

Ten immunoassays changed their response in the presence of high 25(OH)D2 concentrations. Some showed particularly large deviations, while two immunoassays and the LC-MS/MS method were much less affected. This means that assay disagreement is not always a uniform offset that could be corrected with one simple adjustment. Performance can depend on the composition of the sample itself.

This matters especially where vitamin D2 exposure is common. An assay that responds differently to 25(OH)D2 and 25(OH)D3 can create patient-specific bias rather than merely laboratory-wide bias.

Earlier interlaboratory studies showed clinically important misclassification

The modern standardisation effort developed because earlier comparisons showed that analytical differences could change whether a person was labelled low or normal.

Binkley and colleagues reported striking differences between laboratories. In clinically similar groups of postmenopausal women, mean 25(OH)D values differed markedly between two laboratories. In a subsequent experiment, serum from ten healthy adults was divided and sent to six laboratories. Mean measured concentrations ranged from 17.1 to 35.6 ng/mL. When samples were spiked with 20 ng/mL of 25(OH)D, the measured increase ranged from only 7.7 to 18.0 ng/mL.

The implication was direct: classification could depend partly on which laboratory performed the test.

Later work confirmed that the problem was not confined to early assay generations. Comparisons of immunoassays with LC-MS/MS have demonstrated random variability and method-specific bias large enough to change deficiency classification in a substantial proportion of specimens.

Large population studies can inherit assay bias

Analytical noise becomes especially consequential when individual results are aggregated into population prevalence estimates.

Hyppönen and colleagues measured 25(OH)D in more than 7,000 participants from the 1958 British birth cohort using an automated immunoassay and compared a subset against another commercial assay and external quality-assessment data. In 781 directly compared samples, the IDS assay produced concentrations approximately 15.7 nmol/L lower on average than the DiaSorin method. The difference was also substantial relative to the external quality-assessment mean.

Those differences altered estimated prevalence below vitamin D thresholds. Statistical harmonisation improved agreement, demonstrating that some apparent population difference was generated by the measurement system rather than underlying biology.

The same principle was demonstrated in the Irish National Adult Nutrition Survey. Cashman and colleagues applied VDSP standardisation to ELISA-derived values from 1,118 adults and compared the standardised estimates with reanalysis using standardised LC-MS/MS.

Using the original ELISA measurements, the estimated year-round prevalence below 30 nmol/L was 6.5 percent. After applying the standardisation protocol it became 11.4 percent, while full LC-MS/MS reanalysis produced 11.2 percent. At the 50 nmol/L threshold, the corresponding values were 40.0, 43.7 and 45.0 percent.

These are not trivial statistical differences. They illustrate how analytical calibration can alter the apparent scale of low vitamin D status in an entire population.

Standardisation can also change apparent trends over time

When surveys performed in different years use measurement systems that drift or differ in calibration, a false temporal trend can emerge. Retrospective standardisation of national survey data has shown that some apparent changes in average vitamin D status were substantially measurement artefacts.

German national survey data provide another example. When historical 25(OH)D measurements were retrospectively standardised, mean values increased and the estimated prevalence of deficiency fell. The biological population had not been re-measured in the past. What changed was the calibration of the historical measurement data against a common standard.

This is why comparisons between studies require more than checking whether both papers report “25(OH)D.” Assay type, laboratory performance, calibration, standardisation and the measured metabolite composition can materially affect comparability.

Population variation is not the same as assay error

Even after measurement is standardised, population distributions remain broad. People differ in sunlight exposure, latitude, skin pigmentation, diet, supplementation, body size, age and many other determinants of circulating 25(OH)D. These differences are genuine biological and environmental heterogeneity rather than analytical error.

That creates an interpretive problem for population reference ranges. A laboratory reference interval describes a distribution in a selected population. It does not automatically define an optimal biological target, a treatment threshold or the expected concentration for every individual.

Viljoen and colleagues quantified this issue using biological-variation analysis. Between-person variation was estimated at 40.3 percent, while within-person variation was 12.1 percent. The resulting index of individuality was low, approximately 0.3. A low index of individuality indicates that a broad population reference interval may be relatively poor at detecting a meaningful change within one individual.

This does not mean population intervals have no value. It means they answer a different question from serial personal monitoring. A result can remain inside a broad population interval while being substantially different from that person’s previous biological pattern.

Season changes the population distribution itself

Recent reference-interval work reinforces this distinction. A Spanish study using both direct recruitment of healthy volunteers and indirect analysis of laboratory data found higher 25(OH)D distributions during summer than winter. The direct and indirect approaches also produced different reference intervals.

The result does not establish one universally correct vitamin D reference range. Instead, it demonstrates why reference populations, inclusion criteria, season and analytical method must be known before a range is interpreted. This broader issue is explored in Population Reference Ranges Explained.

Why threshold classification is especially vulnerable

A continuous measurement can tolerate modest measurement noise without completely changing its meaning. Threshold classification behaves differently. If a person has a result close to a cut-off, even a small analytical or biological movement can place the result on the other side of the boundary.

This makes apparent categories such as deficient, insufficient or sufficient especially sensitive to assay calibration and sampling context. The problem becomes larger when a threshold developed using one measurement system is applied to results produced by another system without appropriate standardisation.

The correct response is not to disregard thresholds entirely. It is to understand that classification precision is greatest when measurement systems are standardised and the result is interpreted alongside the relevant clinical and physiological context.

LC-MS/MS is not automatically error-free

Liquid chromatography tandem mass spectrometry offers important analytical advantages, including the ability to resolve 25(OH)D2 and 25(OH)D3. However, simply describing a laboratory as using LC-MS/MS does not prove that its results are accurate. Calibration, extraction, internal standards, chromatographic separation, operator performance and traceability still matter.

Standardisation studies have documented variation even among laboratories using LC-MS/MS. The scientifically important distinction is therefore not “immunoassay bad, mass spectrometry good.” The stronger distinction is between methods whose performance and traceability are demonstrated against reference procedures and methods for which they are not.

What this evidence establishes

The evidence establishes that analytical variability can alter individual classification, population prevalence estimates and apparent trends between studies or across time. Different assay systems can have different precision, bias and metabolite recovery. Standardisation substantially improves comparability.

It also establishes that a second layer of variation remains after analytical standardisation: populations themselves differ. Season, geography and participant characteristics alter the distribution of 25(OH)D. Population reference ranges therefore describe a context-dependent distribution rather than a universal biological truth.

What this evidence does not establish

It does not show that every vitamin D result is unreliable. Modern assays can achieve good precision and many meet recognised performance criteria. It does not show that all disagreement between two measurements is analytical error. True within-person biological change can occur. It does not establish that LC-MS/MS is inherently accurate without appropriate calibration and quality control.

It also does not justify creating a new personal target simply because population reference intervals vary. Reference intervals, diagnostic thresholds and biological targets are related but distinct concepts.

Final interpretation

A vitamin D result should be read in two contexts at once: the biological context of the person and the analytical context of the measurement. Assay bias can move a result. Population heterogeneity can broaden the expected distribution. Season can shift both individual and population values. Thresholds can magnify relatively modest changes into different labels. Standardisation reduces one major source of uncertainty, but responsible interpretation still requires recognition of the population and physiological context in which the result was obtained.

For the broader measurement problem, see Variability in Vitamin D Measurements.

Primary sources and publication details

Wise SA, Camara JE, Sempos CT, et al. Vitamin D Standardization Program (VDSP) intralaboratory study for the assessment of 25-hydroxyvitamin D assay variability and bias. Journal of Steroid Biochemistry and Molecular Biology. 2021;212:105917. PMID: 34010687. Read the full study.

Hyppönen E, Turner S, Cumberland P, Power C, Gibb I. Serum 25-hydroxyvitamin D measurement in a large population survey with statistical harmonization of assay variation to an international standard. Journal of Clinical Endocrinology and Metabolism. 2007;92(12):4615-4622. PMID: 17726070. Read the study record.

Cashman KD, Kiely M, Kinsella M, et al. Evaluation of Vitamin D Standardization Program protocols for standardizing serum 25-hydroxyvitamin D data: a case study of the program’s potential for national nutrition and health surveys. American Journal of Clinical Nutrition. 2013. Read the full study.

Binkley N, Krueger D, Cowgill CS, et al. Assay variation confounds the diagnosis of hypovitaminosis D: a call for standardization. Journal of Clinical Endocrinology and Metabolism. 2004;89(7):3152-3157. PMID: 15240586. Read the study record.

Viljoen A, Singh DK, Farrington K, Twomey PJ. Analytical quality goals for 25-vitamin D based on biological variation. Journal of Clinical Laboratory Analysis. 2011;25(2):130-133. PMID: 21438007. Read the full study.

Direct and indirect reference intervals of 25-hydroxyvitamin D: it is not a real vitamin D deficiency pandemic. 2024. Read the full study.

Related reading

Original sources

Source

Read the full study

Wise SA, Camara JE, Sempos CT, et al.

Publication details

Wise SA, Camara JE, Sempos CT, et al. Vitamin D Standardization Program (VDSP) intralaboratory study for the assessment of 25-hydroxyvitamin D assay variability and bias. Journal of Steroid Biochemistry and Molecular Biology. 2021;212:105917. PMID: 34010687. Read the full study.

Read more →
Source

Read the study record

Hyppönen E, Turner S, Cumberland P, Power C, Gibb I.

Publication details

Hyppönen E, Turner S, Cumberland P, Power C, Gibb I. Serum 25-hydroxyvitamin D measurement in a large population survey with statistical harmonization of assay variation to an international standard. Journal of Clinical Endocrinology and Metabolism. 2007;92(12):4615-4622. PMID: 17726070. Read the study record.

Read more →
Source

Read the full study

Cashman KD, Kiely M, Kinsella M, et al.

Publication details

Cashman KD, Kiely M, Kinsella M, et al. Evaluation of Vitamin D Standardization Program protocols for standardizing serum 25-hydroxyvitamin D data: a case study of the program’s potential for national nutrition and health surveys. American Journal of Clinical Nutrition. 2013. Read the full study.

Read more →
Source

Read the study record

Binkley N, Krueger D, Cowgill CS, et al.

Publication details

Binkley N, Krueger D, Cowgill CS, et al. Assay variation confounds the diagnosis of hypovitaminosis D: a call for standardization. Journal of Clinical Endocrinology and Metabolism. 2004;89(7):3152-3157. PMID: 15240586. Read the study record.

Read more →
Source

Read the full study

Viljoen A, Singh DK, Farrington K, Twomey PJ.

Publication details

Viljoen A, Singh DK, Farrington K, Twomey PJ. Analytical quality goals for 25-vitamin D based on biological variation. Journal of Clinical Laboratory Analysis. 2011;25(2):130-133. PMID: 21438007. Read the full study.

Read more →
Source

Read the full study

Direct and indirect reference intervals of 25-hydroxyvitamin D: it is not a real vitamin D deficiency pandemic.

Publication details

Direct and indirect reference intervals of 25-hydroxyvitamin D: it is not a real vitamin D deficiency pandemic. 2024. Read the full study.

Read more →