Another example of a direct plasma melatonin ELISA producing questionable results

A recent paper in this journal[1] has some serious deficiencies with respect to the analysis of melatonin and the conclusions drawn from the results. The primary aims of the study were to investigate the effects of transcutaneous auricular vagal nerve stimulation (taVNS) on dyspepsia symptoms and sleep quality in functional dyspepsia patients with sleep disturbance. Among the secondary aims was the possible involvement of melatonin. The authors collected single samples of venous blood from the subjects in the fasting state, both before and 4 weeks after their intervention, between 0800 h and 1000 h. In the study, taVNS significantly increased serum melatonin from 195 pg/ml at baseline to 253 pg/ml at the end of the 4-week treatment, while melatonin in the taVNS group was significantly higher than that in the tnVNS group (209 pg/ml vs. 191 pg/ml) after 4 weeks. It is widely recognized that a single melatonin measurement (especially in the morning) is not acceptable[2–4]. Melatonin production ceases between 0300 h and 0600 h[5] and plasma levels decrease at a rate determined by individual differences in the half-life of melatonin. The imprecision of the actual sampling times, spanning more than two half-lives, makes any inferences about the timing of the circadian rhythm of melatonin production or even the peak levels of production unreliable. A single collection, even if standardized to ensure that all collections occur at exactly 8 a.m., still provides no meaningful information. The subjects will have varying circadian rhythm timings (some early and some later), different maximal production rates of melatonin, different light exposure intensities and timings the night before, and different half-lives of melatonin metabolism. These results do not reflect the steady-state characteristics of melatonin in the target population. The fact that the authors will be able to cite other published papers using this experimental design does not make their conclusions any sounder. If we now look at the actual melatonin concentrations reported in this paper, there are further concerns. The melatonin levels reported for the morning sampling are at least 20 times higher than expected. There is abundant evidence that when melatonin is assayed using appropriately validated RIA, ELISA, or especially HPLC-MS/MS methods, between 0800 h and 1000 h, the levels of melatonin in plasma are less than 10 pg/ml, not 200 pg/ml[4]. In fact, an assessment of 179 papers in which peak melatonin levels were measured in studies from around the world, using many different methods, found that after 30 years of age, mean melatonin levels were less than 100 pg/ml[6]. Peak saliva melatonin levels are approximately 24 pg/ml, based upon 34 data sets[7]. This situation arises from the fact that the assay used in the study, a direct melatonin ELISA kit from Elabsciences, is inaccurate and grossly overestimates the levels of melatonin. The ELISA kit does not state that it has been validated but claims that “it applies to the in vitro quantitative determination of human melatonin concentrations in serum, plasma, and other biological fluids.” The kit manufacturers do not provide any evidence that the kit detects only melatonin or details of which potential cross-reacting compounds were actually tested. The assay relies entirely on the specificity of the antibody used, as there is no extraction/purification step. The testing by the company for matrix effects, which is an important issue with direct assays, is not reported in the kit protocol document. Again, the authors will be able to find papers that have used this ELISA kit, but a common feature will be that they report extremely high melatonin levels at times of the day when the pineal gland is not producing it. The gastro intestinal tract does not contribute to plasma melatonin levels[8]. One paper on mice that used the Elabscience ELISA kit actually reported levels of 370 pg/ml in the C57Bl/6 strain of mouse that is genetically incapable of synthesizing melatonin[9]. A recent paper that used the Elabscience ELISA in humans has removed all melatonin results in a published correction because “the assay for melatonin may have been invalid for these purposes and as such all the melatonin data shown in the abstract and results sections should be ignored”[10]. Unfortunately, there are several companies, like Elabsciences, that are selling direct serum melatonin ELISA kits that produce grossly abnormal high melatonin results. Many reviewers do not have the expertise to raise questions about assay accuracy during the review process, and journals can be reluctant to retract or correct published papers, but it is happening. The following have been retracted because of melatonin assay issues:[11–14] In science, we strive for accuracy. If we say that we have detected differences in melatonin levels, then we must be confident that the values are accurate. I cannot conceive of seeing a published study that reports insulin or troponin levels that are 10 or 100 times the expected concentration. Just because authors can find other papers that also report extremely high melatonin levels at 0900 h does not make them correct. The authors concluded that taVNS, but not tnVNS, increased serum melatonin levels, and taVNS enhanced vagal activity, with HF in the fasting state positively correlating with serum melatonin levels. We can have no confidence that their treatment affected melatonin levels. Indeed, we can have no confidence that the assay they used actually measured melatonin specifically and not other substances as well, or that it was not subject to matrix effects. Ethical approval Not applicable. Consent Not applicable.