A plasma concentration looks like the most solid kind of number in a research paper. It has units. It came out of an instrument. It reads like a property of the sample.
It is a property of the sample and of the assay that read it. This site has made the same argument three times already about different numbers: an HPLC purity percentage belongs to its method, an EC50 belongs to its cell system, a half-life belongs to its route and species. Measured concentrations in blood, plasma or serum are the fourth member of that family, and the one most often quoted as if it were universal. This piece is about how they are produced and what must accompany one for it to mean anything.
Bind it or weigh it — and what an antibody cannot tell apart
Essentially every reported peptide concentration comes from one of two approaches.
Immunoassay uses antibodies to capture the analyte — radioimmunoassay, ELISA, chemiluminescence and electrochemiluminescence variants all work this way. A sandwich format uses two antibodies binding two different sites, so both have to be present and intact for a molecule to be counted. The critical fact is structural: an antibody does not recognise a molecule, it recognises an epitope — a small region of one. Anything carrying that region gets counted; anything missing it does not, however similar the rest of the molecule is.
LC-MS/MS separates the sample chromatographically, then identifies and quantifies by mass and fragmentation, normally against a stable-isotope-labelled version of the analyte added as an internal standard. Specificity comes from retention time plus mass transitions rather than from molecular recognition.
So an immunoassay reports how much material its antibody pair could bind, while a mass spectrometer reports how much of a species with a particular mass and fragmentation pattern was present. Neither is generically superior, and their failure modes do not overlap.
The cleanest demonstration is in the incretin literature. Bak et al., Diabetes, Obesity and Metabolism 2014;16:1155-1164 (PMID 25041349) took ten commercial GLP-1 kits, spiked six synthetic GLP-1 isoforms into each kit's own buffer and into human plasma, and measured recovery.
The results were not small differences in calibration. One kit detected none of the six isoforms. The "active" ELISAs were specific for intact GLP-1. One total-GLP-1 kit detected all six forms in buffer but mainly amidated forms in plasma. Others detected only amidated species, or had a low measurement ceiling, or a plasma detection limit sitting above much of the physiological range. The authors' summary of the clinical samples is the line worth memorising: the pattern of postprandial responses was similar between kits, but the absolute concentrations measured varied.
That is the mechanism behind a point made in this site's incretin axis explainer — that "total" and "active" GLP-1 are different measurements. DPP-4 converts most circulating GLP-1(7-36)amide to GLP-1(9-36)amide by removing two N-terminal residues. An assay whose antibodies require the intact N-terminus counts only the uncleaved form; an assay reading a C-terminal epitope counts both. Neither is wrong — they answer different questions in the same units. When the metabolite itself became interesting, a dedicated assay had to be built: Wewer Albrechtsen et al., American Journal of Physiology — Endocrinology and Metabolism 2017;313:E284-E291 (PMID 28420649) developed a sandwich ELISA recognising GLP-1(9-36)amide, stating plainly that reliable assays were not available.
"Plasma GLP-1 was X pmol/l" is therefore incomplete in exactly the way "purity 98.5%" is incomplete. The number needs its method attached.
The sample changes before anyone measures it
Peptides do not wait politely in the tube. The peptidases that set clearance in a living system are present in drawn blood, and they keep working.
Wewer Albrechtsen, Danish Medical Journal 2017;64(11):B5425 (PMID 29115211), reviewing work on GLP-1, glucagon and oxyntomodulin, reports that inappropriate sample handling may cause up to 50% variation in results. Collection additives, temperature, time before freezing and freeze-thaw history are all inputs to the final number, and are almost never visible in an abstract.
The mass spectrometry side has its own version. Matuszewski et al., Analytical Chemistry 2003;75:3019-3030 (PMID 12964746) challenged the then-common assumption that LC-MS/MS is inherently selective, describing how co-eluting matrix components cause ion suppression or enhancement — the signal changes size because of what else is in the sample, not because of how much analyte is there. MS resolves identity problems that defeat antibodies; it does not resolve quantitative ones by itself.
When two methods disagree: the IGF-1 case
IGF-1 is the most instructive example on this shelf, because it is measured routinely, standardisation has been worked on for decades, and it is still hard.
Bonert et al., Pituitary 2018;21:65-75 (PMID 29218459) prospectively compared an established immunoassay against a consensus-compliant LC-MS method in 101 patients with pituitary disease. Twenty-four percent of samples were classified differently as below, within, or above the reference range, agreement was poor overall, and the immunoassay was biased toward higher values.
Lee et al., Clinica Chimica Acta 2023;539:130-133 (PMID 36528048) ran 110 sera across three automated immunoassay platforms and LC-MS/MS. Weighted Deming regression put pairwise differences between roughly 8% and 81% depending on which two methods were compared, agreement was poor in a way reflecting systematic difference rather than noise, and the spread widened as IGF-1 values rose.
Part of why IGF-1 is difficult is a fact covered from the pharmacology side in the IGF-1 LR3 deep dive: most circulating IGF-1 is held in complexes with IGF-binding proteins at affinities at or above the receptor's. An assay has to liberate the analyte from those complexes first, and how completely it does so is a property of the method.
The reading consequence is narrow and firm. Within one study on one platform, an IGF-1 comparison is sound. Across two studies on two platforms, the absolute values are not interchangeable.
Measuring an analog is a different problem from measuring the hormone
Assays for endogenous peptides are built against endogenous epitopes. An engineered analog may or may not be recognised, and the direction of the error is not predictable from structure.
Luo et al., Analytical Biochemistry 2026;714:116115 (PMID 41871707) shows what closing that gap requires. To quantify semaglutide in human serum the group raised a pair of anti-semaglutide monoclonal antibodies, built an electrochemiluminescence immunoassay around them, validated it, and demonstrated no interference from recombinant human GLP-1 — engineered so the endogenous counterpart does not register. Read in reverse: a kit for native GLP-1 is not a kit for semaglutide, and making one meant new reagents.
Antibody specificity is not only a hazard, though — sometimes it is the whole method. Recombinant somatropin, the catalog's HGH 191AA, is a single molecular form, while the pituitary secretes a mixture. Anti-doping laboratories exploit that: one assay weighted toward the full-length 22 kDa form, a second recognising multiple pituitary forms, and the ratio between them as the readout. Marchand et al., Drug Testing and Analysis 2022;14:724-732 (PMID 34761559) consolidated that pair into one duplex microplate assay, and is candid that its ratios ran lower than the established immunoluminometric assays', so decision limits would need re-deriving.
For most compounds on a research shelf, something more basic applies: no validated bioanalytical method for them exists in the public literature. That absence is part of why half-life figures circulate for MGF and PEG-MGF or the short TB-500 heptapeptide that were never measured in anyone.
"Levels decline with age" is a measurement claim first
This site has repeatedly flagged decline-with-age framings — for NAD+, GHK, the mitochondrial-derived peptides — because a falling biomarker may be cause, consequence, or proxy. Assay literacy adds a prior step: before it is a biology claim, it is a claim about what an assay read.
Exp Physiol 2026;111:3772-3783 (PMID 42349896), from Karolinska with the USC group that founded the field, measured humanin, MOTS-c and SHMOOSE with an in-house ELISA in adults with cerebral palsy and typically developing controls. Despite lower muscle mass and peak oxygen uptake, the cerebral palsy group showed basal MDP levels comparable to controls, with no or only modest changes after exercise. In-house is not a criticism — no standardised commercial reference method exists for these analytes — but it ties the values to one laboratory's antibodies and calibrators. And the finding is a null.
Two more 2026 examples show how much can rest on a single measured number. In Archives of Endocrinology and Metabolism (online 7 April 2026, PMID 41945630), serum MOTS-c by ELISA in 121 adolescents with PCOS versus 125 controls was higher in the PCOS group but did not reach significance (p = 0.059), with no association to metabolic parameters — and every participant carried the wild-type genotype at the m.1382A>C position discussed in this site's mitochondrial peptide roundup. And Diagnostics 2026;16(13):1979 (PMID 42449760) is a pilot whose whole design rests on sorting fasting GLP-1 and GIP immunoassay values into tertiles; the authors call it hypothesis-generating and name among their own critical limitations the inability to distinguish incretin secretory deficiency from receptor resistance using fasting measurements alone.
Five questions to ask of a reported concentration
- Which method? If immunoassay, which antibody pair or kit — and is it reading total or intact analyte?
- Which matrix, handled how? Serum or plasma, what additives, how long before freezing.
- Endogenous peptide or analog, and was the method validated for the thing actually measured?
- Absolute value, or within-study pattern? Bak et al. found the pattern reproduced across kits while the absolute numbers did not.
- Is it comparable to the other paper's number? Usually not, unless the methods match.
None of this is a documentation question. No certificate of analysis line, chromatogram or mass spectrum reports how much of a compound was present in a biological sample — like cell entry, this is a property of an experiment, not of material in a vial. See /quality/ and the /library/.
FAQ
Is LC-MS/MS simply the better method? It is the better method for resolving which species is present, which is why it wins arguments about isoforms and metabolites. But it needs an appropriate internal standard, it is vulnerable to matrix effects, and it stays blind to distinctions carrying no mass difference, as covered in the chiral purity piece. The useful question is which method suits the analyte.
Why do research peptides not have reference ranges? A reference range requires a defined population measured by one standardised method against a common calibrator; for most compounds here, neither exists. An in-house assay produces internally consistent values that are not transferable.
Does any of this change how I read a COA? Not directly, and that is the point. Concentrations measured in biological samples sit with replication, delivery and cell entry: established by experiment, reported in papers, and absent from product documentation by nature rather than omission. Evaluating them belongs to the evidence hierarchy, not the paperwork.
This article is educational and for the laboratory research community. Trulogic Labs products are sold for laboratory and research use only and are not for human consumption.