GDF15

In brief

What GDF15 is, and how one gene makes two different things

Growth differentiation factor 15 is a hormone of distress. Cells make almost none of it when they are comfortable, and a great deal of it when they are not: when their mitochondria falter, when they are short of oxygen, when a drug poisons them, when a pathogen injures them, when they turn cancerous, and, as a normal part of life, when a placenta is growing ^5,29,30,31. The protein leaves the cell, travels in blood, and acts on one target, a receptor in the floor of the fourth ventricle at the base of the brain ^6,8. The message it delivers is roughly: stop eating, and feel sick ³².

Structurally it belongs to the transforming growth factor beta superfamily, a very large group of signaling proteins that control growth, repair, immunity and the shaping of embryos. The group that purified it in 1997 noted that it had no strong homology to any existing family within that superfamily and might be the first of a new one; they named it macrophage inhibitory cytokine 1 because purified recombinant protein suppressed the production of tumor necrosis factor alpha by macrophages treated with bacterial endotoxin ⁵. That first paper also established the shape of the molecule: expressed in hamster ovary cells, the precursor had its propeptide cleaved off and the cell secreted a cysteine-rich dimer of about 25 kilodaltons ⁵.

The gene is simple. A parallel 1997 paper mapped it to the short arm of chromosome 19 by fluorescence in situ hybridization and found a single-copy gene of only two exons, of 309 and 891 base pairs, separated by one intron of about 2.9 kilobases; the main messenger RNA is 1.2 kilobases, and a second transcript of 1.9 kilobases predominates in adult skeletal muscle ¹. That same paper reported a precursor protein of 295 amino acids and a mature protein of 112; a third 1997 paper, published three months earlier, in September, read the same gene as encoding 308 amino acids ^1,3. The discrepancy is in the choice of start codon, not in the biology, and the mature 112-amino-acid chain is the same in both accounts.

One feature of the protein matters more than it sounds. A cell can secrete GDF15 either as the finished mature dimer or with the propeptide still attached, and only the unprocessed form sticks to the extracellular matrix, the protein scaffolding between cells; the sticking is done by the last 89 amino acids of the propeptide ³³. So the tissue holds a latent store that the blood test never sees. When the propeptide was deleted in a mouse tumor graft, serum levels rose about twentyfold, and in human prostate cancer tissue the amount held in the stroma, rather than the amount circulating, independently predicted relapse after the prostate was removed ³³. A later study of monocytic cells showed that the unprocessed precursor is secreted fast and the mature protein slowly, probably by a different route, so the ratio between the two is itself a control point ²⁹.

Cloned seven times in five years, under six names

GDF15 has an unusual history: between 1997 and 2001 seven groups, chasing unrelated questions, cloned or characterized the same gene, and six different names were attached to it ^{1,3,4,5,10,11,12}. Four of those groups named it in 1997 and 1998, each differently; one reported it without naming it at all ³; the name that stuck, GDF-15, was coined in 1999 by a group in Heidelberg ¹²; and a sixth name arrived in 2001, from the first group to notice the pile-up ¹¹. Nobody was wrong, and for several years nobody noticed they were describing one protein.

In Sydney, a group looking for genes switched on when macrophages are activated found it by subtraction cloning and called it macrophage inhibitory cytokine 1, MIC-1, because the purified protein suppressed macrophage tumor necrosis factor production ⁵. In Japan, a group testing a new method for spotting secreted proteins pulled the same complementary DNA out of a library and reported it simply as a novel superfamily protein highly expressed in placenta ³. In New York, a placental gene-discovery project named it placental transforming growth factor beta, PTGFB, and worked out the gene structure and chromosomal location ¹. In Indianapolis, a group studying bone morphogenetic proteins isolated it from placenta, judged it closest to BMP-8, named it PLAB, and showed that it inhibited the growth of primitive blood-forming progenitor cells ¹⁰. At Pfizer in Connecticut, a group looking for new bone morphogenetic proteins found it expressed in placenta and prostate, named it prostate-derived factor, noted that it carries two cysteines in addition to the seven that mark the superfamily, showed that androgens control it in the rat prostate, and reported that implanting the recombinant protein under the skin produced cartilage and the beginnings of bone ⁴.

The name the gene now carries came from none of those groups. In 1999 a group in Heidelberg screened genomic libraries and isolated the rat, mouse and human genes together, found all three built of two exons with a single intron interrupting the coding sequence at the identical position within the prepro-domain, and labeled the protein growth/differentiation factor-15, GDF-15, alongside MIC-1. The same paper reported that the rat, mouse and human versions of this protein show the lowest sequence conservation of any member of the transforming growth factor beta superfamily, which is a standing warning about reading mouse results across to people ¹².

The sixth name came from a different direction. Researchers at the National Institute of Environmental Health Sciences, trying to explain why nonsteroidal anti-inflammatory drugs reduce colon cancer risk in ways their known target, cyclooxygenase, could not account for, used subtraction cloning on drug-treated colon cancer cells and found a gene they called NSAID-activated gene 1, NAG-1. Their 2001 paper opens by observing that their new gene has an identical sequence to a protein that already has five different names, which made NAG-1 the sixth ¹¹. Cells forced to express it were more prone to programmed cell death and formed fewer colonies in soft agar, and tumors grown from them in mice without immune systems grew less well ¹¹.

Two things follow from this scattered beginning. First, the early literature on GDF15 is split across six keywords, which is part of why it took twenty years to find the receptor. Second, every one of those first descriptions was about something other than appetite: macrophages, placenta, bone, prostate, colon cancer, comparative genetics ^{1,3,4,5,10,11,12}. The function the hormone is now defined by was not on anyone's list.

What turns it on, and where it is made

There is no single organ of GDF15. Almost any tissue will make it under enough strain, which is why the blood level behaves like a thermometer for systemic stress rather than a readout of one gland. The macrophage was the first source described, and in monocytic cells the gene answers to interleukin-1 beta, tumor necrosis factor alpha, interleukin-2, macrophage colony-stimulating factor and transforming growth factor beta itself, but not to interferon gamma and not to bacterial lipopolysaccharide ⁵. It also answers to oxidized low-density lipoprotein, and the protein is abundant in the macrophages inside atherosclerotic plaques ²⁹. Infection raises it from the organ that is infected. In patients with acute respiratory failure, plasma GDF15 tracked the levels measured in the lower respiratory tract and was higher in those who died; severe acute respiratory syndrome coronavirus 2 induced GDF15 expression in human lung epithelium, and lower respiratory tract levels were higher in the patients with coronavirus disease who died. In mice, deleting the gene from the lung epithelium alone blunted the release into the airspace and the blood caused by exoproducts of the type II secretion system of Pseudomonas aeruginosa, and putting GDF15 back into the airspace raised circulating red cell counts rather than changing the local cytokine profile ³¹. The human half of that work is observational and prognostic; the mechanism is mouse.

Inside the cell the most important switch is the integrated stress response, the emergency program that runs when protein folding, amino acid supply or mitochondrial function goes wrong. In mouse liver cells metformin turns on GDF15 secretion by raising ATF4 and CHOP, short for activating transcription factor 4 and C/EBP-homologous protein, two transcription factors of that program ³⁰. GDF15 is also a major target of p53, the tumor suppressor that senses DNA damage, which links it to the cancer literature from the other side ²⁹. And it rises with mitochondrial disease severity in direct proportion to a measurable biochemical derangement: in 102 patients with the m.3243A>G mutation, nearly all of the twenty blood markers that separated patients from controls, GDF-15 among them, were attributable to an increased ratio of NADH to NAD+, the reduced and oxidized forms of the cell's main electron carrier ⁴⁷.

The organ sources that matter for specific situations have been worked out one at a time, mostly in rodents, and the answers differ by stimulus. For metformin, deleting the gene from the intestinal epithelium of mice markedly reduced the rise in blood GDF15 after an oral dose, while deleting it from the liver did nothing; for a direct activator of AMP-activated protein kinase, the enzyme that senses when a cell's energy currency is running down, the pattern reversed and the liver mattered ⁴⁸. Rats given metformin while eating a high-fat diet raised GDF15 chiefly from the kidney, and knocking the gene down in the kidney, or knocking the receptor down in the area postrema, abolished the effect on feeding ⁴⁹. Fat delivered into the upper small intestine of rats also raises kidney GDF15 and suppresses eating, through the same kidney-to-brainstem route, and high-fat feeding breaks that link ⁵⁰. One human experiment makes the gut source concrete. In a placebo-controlled, double-blind, randomized two-visit crossover in healthy volunteers, a single oral dose of hydroxychloroquine sulfate, a bitter-tasting drug, raised plasma GDF15, and the rise correlated with lower hunger scores and lower plasma ghrelin, the stomach's hunger hormone. In jejunal crypt cultures from patients with obesity, GDF15 was expressed in hormone-secreting and mucus-secreting cells, more so than in lean tissue, and various bitter compounds raised or lowered its expression through particular subtypes of bitter taste receptor and through the unfolded protein response, the branch of the stress response that deals with misfolded protein; the bitter antibiotic azithromycin acted instead through the motilin receptor, which the authors suggest explains some of its aversive side effects ⁵¹. The crossover was small, and a correlation with hunger scores is not a weight outcome.

In pregnancy the source is not the mother at all. Using a naturally occurring labeled variant of the protein and mass spectrometry to tell maternal from fetal molecules apart, a 2023 study showed that the vast majority of GDF15 in maternal plasma comes from the feto-placental unit ²¹. Single-cell work in the 2026 genetic study localized the expression to extravillous trophoblasts, the placental cells that invade the lining of the womb ⁴⁶.

Ordinary life moves the number too. In 228 people aged 18 to 83, plasma GDF15 rose with age and was lower in habitually active cyclists than in healthy controls recruited from the general population at any age, but a single long-distance race raised it sharply, to the levels seen in people immobilized by lower-limb disease ⁵². Across 36 human proteomic studies of aging, GDF15 is one of only 32 proteins reported to change with age by five or more independent analyses ⁵³.

Twenty years without a receptor, then four groups in one month

For two decades GDF15 was a protein whose effects were measurable and whose mechanism was not. The 2007 paper that established its role in wasting proposed that it acted on the hypothalamus through the type II transforming growth factor beta receptor, with signaling through the extracellular signal-regulated kinases 1 and 2 and the signal transducer and activator of transcription 3, and effects on neuropeptide Y and pro-opiomelanocortin, two of the hypothalamus's appetite-setting signals ³⁸. That account is no longer held. Part of the reason is a 2017 paper from Trondheim showing that batches of purified recombinant GDF15 produced in mammalian cells vary in purity and can carry transforming growth factor beta, which is active at femtomolar concentrations; when the authors tried to characterize GDF15 receptors, the effects they measured were caused by the contaminant, and they wrote plainly that some already published conclusions about GDF15 may not be supported by the data ⁴¹.

The real receptor was found in 2017 by four pharmaceutical groups working in parallel and, remarkably, publishing within a month. Janssen, Eli Lilly and Novo Nordisk published on the same day in Nature Medicine, and NGM Biopharmaceuticals published a month later in Nature ^6,7,8,20. All four identified GFRAL, which stands for glial cell line-derived neurotrophic factor family receptor alpha-like, an orphan member of a receptor family previously associated with neurotrophic factors. GDF15 binds GFRAL with high affinity; GFRAL alone cannot signal and must partner with RET, a receptor tyrosine kinase ^6,8. Mice lacking GFRAL do not lose weight or eat less when given GDF15, and an antibody that blocks the interaction prevents the metabolic effects in rats ^6,7.

The anatomy is the striking part. GFRAL messenger RNA was mapped to neurons of the area postrema and the nucleus of the solitary tract, in mouse, rat, monkey and human, and the Novo Nordisk group reported no expression in peripheral tissues at all ^6,7,8. The area postrema is one of the few places where the blood-brain barrier is deliberately incomplete; it exists to sample the blood for poisons. A hormone whose only known receptor is there is, by construction, a message about toxicity. That mapping was done by messenger RNA detection in 2017, and absence of a detected transcript is weaker evidence than presence of one; later reports of low-level GFRAL expression outside the hindbrain are collected in a 2025 review, which treats a broader distribution as a live possibility, and as one proposed explanation for the effects of GDF15 that do not look like brainstem effects, rather than as a finding ⁹.

The NGM group traced the circuit one step further: GDF15 activates GFRAL neurons in the area postrema and solitary nucleus, which activate neurons in the parabrachial nucleus and the central amygdala, the pathway they call the emergency circuit for feeding responses to stress. Their knockout mice ate more under stressed conditions and were resistant to the appetite loss and weight loss caused by chemotherapy ²⁰.

In 2019 cryo-electron microscopy settled how the three proteins fit together. Structures of the extracellular GDF15-GFRAL-RET complex, alongside three related complexes, showed that each ligand and co-receptor pair, despite using different atomic contacts, forces RET into the same dimeric arrangement, poised so that the two intracellular kinase domains can phosphorylate each other ³⁴.

The most recent work on these neurons complicates the idea that they exist only to report poisoning, and it is a preprint, not yet peer reviewed. Recording optically from area postrema neurons in behaving mice, the authors found that GFRAL neurons are activated when the animal eats food rich in fat, that this fat-specific activation is required for the animal to stop eating fat, and that it does not require GDF15 or the usual gut-to-brain pathways; neighboring neurons carrying the receptor for glucose-dependent insulinotropic polypeptide, which directly inhibit GFRAL neurons, were activated instead by sugar. Both cell types are targets of the weight-loss drug tirzepatide. If it holds, GFRAL is a nutrient sensor that GDF15 can hijack rather than a dedicated poison alarm. It is mouse work, and it is pending review ⁵⁴.

Not loss of appetite: nausea

Early work described GDF15 as an anorectic hormone, a molecule that reduces food intake, and the distinction between not being hungry and feeling sick was not pressed. In mice, overexpressing it produced a lean animal that ate less spontaneously, with unaltered energy expenditure and improved glucose tolerance, which reads as a clean appetite effect ⁵⁵. The lesion studies narrowed the site: injected into mice, GDF15 lit up neurons in the area postrema and the medial nucleus of the solitary tract, and destroying both structures completely abolished its effect on feeding ³⁵.

The reframing came in 2020. Giving GDF15 to musk shrews, Suncus murinus, a small mammal that can vomit, caused vomiting. Giving it to mice and to lean and obese rats, which cannot vomit, caused the behaviors those species use instead, including pica, the eating of inedible material, which rodent researchers use as a proxy for nausea. The same study showed that GDF15 contributes to the malaise mice develop after chemotherapy ³². A review of the field four years later, published online in 2024, concluded that the anorexia is downstream of nausea and emetic circuitry, not a separate appetite action ³⁶.

The clinical corollary arrived from the chemotherapy side. Platinum drugs raise circulating GDF-15 in people, and in a colorectal cancer study the rise was positively associated with weight loss; in mice, the chemotherapy agents with the highest clinical emetic scores were the ones that raised GDF-15 most. Mice lacking GDF-15 had less anorexia and weight loss after platinum chemotherapy, a neutralizing antibody improved their survival, and in nonhuman primates the antibody reduced both appetite loss and vomiting ²⁷.

One part of the 2017 account has since been qualified rather than overturned. The Novo Nordisk paper found that GDF15 did not raise energy expenditure in obese mice, and concluded that the anti-obesity action was therefore likely to be driven primarily by reduced food intake ⁸. In 2023 a mouse study with Novo Nordisk coauthors reported that during caloric restriction recombinant GDF15 also prevented the compensatory fall in energy expenditure, the adaptation that makes lost weight come back, producing greater weight loss and less fatty liver than caloric restriction alone; the effect required GFRAL and a beta-adrenergic signal that raised fatty acid oxidation and calcium futile cycling in skeletal muscle ⁴⁴. The Cambridge metformin work had already reported that the drug's effects on both energy intake and energy expenditure depended on GDF15 ⁴². All of this is mouse physiology, and no human study has measured energy expenditure on a GFRAL agonist.

This matters for how the molecule is understood. A hormone that makes animals vomit ³², released by damaged tissue, acting on the one brain region built to detect poisons ^6,8, is best read as part of an ancient defense: something is wrong with what is inside you, so stop taking more in. Whether that defense is helpful in a person dying of cancer is a separate question, and the trials in the next section are the attempt to answer it.

Cancer cachexia, and the first drug aimed at the hormone

Cachexia is the wasting of advanced illness: weight falls, muscle disappears, appetite goes, and function and survival go with them. It affects up to 80 percent of patients with advanced cancer, and as of a 2025 review no drug is approved for it by either the United States or the European regulator ⁵⁶. The benchmark for drug effects in cancer cachexia before GDF15 was anamorelin, a ghrelin-receptor agonist: in two randomized, double-blind phase 3 trials in 484 and 495 patients with stage III or IV non-small-cell lung cancer and cachexia, median lean body mass over 12 weeks rose 0.99 kg, 95 percent confidence interval 0.61 to 1.36, against minus 0.47 kg, minus 1.00 to 0.21, on placebo in the first trial, and 0.65 kg, 0.38 to 0.91, against minus 0.98 kg, minus 1.49 to minus 0.41, in the second, both with p below 0.0001, while handgrip strength did not differ in either trial. Those trials were funded by Helsinn Therapeutics, and lean body mass is not survival ²². The GDF15 hypothesis for cachexia began in 2007, when serum MIC-1 in men with advanced prostate cancer was found to track their weight loss, mice bearing MIC-1-secreting tumors wasted, an antibody against the protein reversed the wasting, and transgenic mice overexpressing it ate less and weighed less ³⁸.

The first patients were dosed in an open-label phase 1b study of 10 adults with advanced cancer, cachexia and a raised GDF-15, given ponsegromab 200 mg under the skin every three weeks for 12 weeks alongside their cancer treatment. Median unbound serum GDF-15 was 2.269 ng/mL at baseline and fell below the assay's lower limit of quantification, 0.0424 ng/mL, from three hours after the first dose through week 15; the least-squares mean body weight rose 4.63 kg, standard error 1.98, by week 12, about 6.6 percent of baseline; no treatment-related adverse events were identified. Ten patients, open label, no control group, and funded by Pfizer ⁴⁵. Seventeen years after the hypothesis was framed, it was then tested properly. Ponsegromab is a humanized monoclonal antibody that binds and neutralizes GDF-15. In a randomized, double-blind, 12-week phase 2 trial funded by Pfizer, 187 patients with cancer cachexia and serum GDF-15 of at least 1,500 pg/mL received 100, 200 or 400 mg every four weeks, or placebo; 40 percent had non-small-cell lung cancer, 32 percent pancreatic cancer and 29 percent colorectal cancer. The median difference in weight change against placebo at 12 weeks was 1.22 kg at 100 mg, 95 percent credible interval 0.37 to 2.25, 1.92 kg at 200 mg, 0.92 to 2.97, and 2.81 kg at 400 mg, 1.55 to 4.08, so the interval excluded zero at all three doses. Appetite, cachexia symptoms and measured physical activity improved at the top dose, and adverse events of any cause occurred in 70 percent on drug and 80 percent on placebo ¹⁵. The trial was preregistered with a prespecified Bayesian analysis of weight as the primary endpoint ⁵⁷.

Three cautions are worth stating alongside that result. The trial lasted 12 weeks and measured weight, not survival; it enrolled only patients with high GDF-15, so it does not describe cachexia in general; and it was designed, funded and analyzed by the company that owns the drug, with most authors employed by it ^15,57. A 2025 review places ponsegromab at the head of the pipeline and entering phase 3, with other antibodies, AV-380 and NGM120, behind it ⁵⁶; a 2026 review co-written by one of the trial's own investigators names the GDF-15 to GFRAL axis as a primary mediator of cachexia and lists further anti-GDF-15 and anti-GFRAL agents in development, while warning that trial design and endpoints remain the field's weak point ⁵⁸. Company work in mice suggests that combining GDF-15 blockade with an anti-myostatin antibody or the appetite stimulant anamorelin adds something only in tumors with low GDF-15 to begin with ⁵⁹. The trial also drew published correspondence in the New England Journal of Medicine in March 2025, with a reply from the Pfizer authors; the PubMed records for those letters carry no abstract, so their substance cannot be quoted here, only the fact that the result was contested in print ^60,61,62. As of October 2026 no phase 3 trial of ponsegromab has reported.

A dissenting line of argument deserves airtime. Two gerontologists writing in 2026 point out that GDF15 is the signal by which the body reports energetic stress to the brain and reallocates resources; suppressing it should relieve symptoms, and may do so at the cost of whatever that reallocation was accomplishing. Their piece is a viewpoint, not data, and the question it raises, whether people who gain weight on anti-GDF15 drugs live longer or shorter, is exactly what a phase 3 trial with survival endpoints could answer ⁶³.

A separate oncology use has emerged from the tumor's side rather than the patient's. Tumors that produce GDF-15 appear to suppress the immune attack against themselves. In the first-in-human GDFATHER-01 phase 1/2a trial, patients whose cancers had stopped responding to immunotherapy that blocks programmed cell death protein 1 (PD-1) or its partner molecule PD-L1, the brake a tumor uses to switch off the T cells attacking it, were given visugromab, a neutralizing anti-GDF-15 antibody, at 10 mg/kg with nivolumab 240 mg every two weeks; durable deep responses occurred in some patients with non-squamous lung cancer and urothelial cancer, the two tumor types in which an in silico screen of about 10,000 tumor samples from The Cancer Genome Atlas most often linked the tumor's own GDF-15 to a suppressed immune response, and treated tumors showed more cytotoxic T cell infiltration, proliferation, granzyme B expression and interferon-gamma signaling ^16,64. That paper carries a published author correction ⁶⁵. Long-term follow-up of the 77 patients in the three expansion cohorts reported objective response rates of 18.2 percent in 22 lung cancers, 18.5 percent in 27 urothelial cancers and 14.3 percent in 28 liver cancers, with median response durations between 19 and 32 months ⁶⁴. Among the 13 responders, 8 had a confirmed complete response or complete metabolic response, 7 of those still ongoing at the data cut, and 6 responded more deeply than they had to their prior checkpoint antibody, with a median response duration of 28.8 months, 95 percent confidence interval 7.4 to 38.0, against 12.0 months, 8.0 to 24.0, on that earlier treatment ⁶⁴. That last comparison is within the same patients and not randomized, so it cannot separate the drug from the selection of patients who were able to respond at all. These are single-arm results in heavily pretreated patients; they are encouraging and they are not controlled, and the authors say so ⁶⁴. A narrative review of GDF15 across solid tumors finds high levels associated with worse prognosis in colorectal, gastric, pancreatic, breast, lung, prostate and head and neck cancers, with laboratory evidence for effects on metastasis and chemoresistance; association is the right word for most of that literature ⁶⁶.

Metformin: a clean story, then an argument

Metformin is the most prescribed drug for type 2 diabetes and is also used to prevent it; more than 60 percent of its preventive effect is attributable to the weight it takes off, and for sixty years nobody knew how it did that ⁴². In late 2019 two groups published, within weeks of each other, the same answer: metformin raises circulating GDF15, and GDF15 is how metformin lowers body weight.

The Cambridge group showed in two independent randomized controlled trials that metformin raises circulating GDF15 in people. In wild-type mice, oral metformin raised circulating GDF15 with expression rising mainly in the distal intestine and the kidney; metformin prevented weight gain on a high-fat diet in normal mice but not in mice lacking GDF15 or lacking GFRAL; in already obese mice an antibody that blocks GFRAL reversed the weight effect; and the drug's glucose-lowering action survived intact without GDF15, showing the two effects are separable ⁴². The McMaster group, working from the liver, showed that metformin induces GDF15 in mouse hepatocytes through ATF4 and CHOP, that the effects on food intake, body mass and fasting insulin vanish in GDF15-null mice, and that rising serum GDF15 was associated with weight loss in patients with type 2 diabetes taking the drug ³⁰.

In 2022 a Copenhagen group tried to build on this and could not. They confirmed, in both mice and humans, that metformin raises GDF15. But in their hands metformin still lowered body weight in GDF15-knockout and in GFRAL-knockout mice, and they concluded that the pathway is dispensable for metformin's effect on energy balance, questioning whether the drug is a strong enough stimulus to drive anorexia through this route ⁴³. This is the kind of disagreement that does not get resolved by one side being careless; both groups used knockout mice and both reported what they saw.

The organ-source work since then has made the picture more specific rather than more settled. In mice, deleting GDF15 from the intestinal epithelium blunted the rise after oral metformin while deleting it from the liver did not, and yet intestinal GDF15 turned out not to be required for GDF15's restraint of high-fat-diet weight gain ⁴⁸. In high-fat-fed rats, metformin raised GDF15 mainly from the kidney, and either knocking the gene down in the kidney or knocking GFRAL down in the area postrema abolished the feeding effect ⁴⁹. The same Toronto group then showed that fat infused into the upper small intestine uses the same kidney-to-area-postrema circuit to suppress eating in chow-fed rats, and that high-fat feeding breaks it ⁵⁰.

The best recent human data are mechanistic rather than decisive. In a randomized crossover study, 34 healthy people fasted for 42 hours twice, once after a week of metformin. Metformin raised serum GDF15 roughly 1.7-fold, from 607 plus or minus 89 to 1,004 plus or minus 61, while FGF21, a liver-derived stress hormone that often travels with it, did not move; the change in GDF15 did not correlate with the metformin concentration in plasma; and human intestinal cells in culture secreted GDF15 and shifted toward glycolysis when exposed to the drug. The abstract prints these values in ng/mL, which for serum GDF15 must be a misprint for pg/mL ⁶⁷. Two trials in girls point the same way from a different angle: four years of metformin in girls born small who had gained weight rapidly raised GDF15 only in years three and four, and the rise was inversely related to insulin, androgens and visceral fat ⁶⁸; and in adolescent girls with polycystic ovary syndrome, a low-dose combination containing metformin raised GDF15 3.4-fold, from 308 to 1,045 pg/mL, while lowering inflammation and liver fat without changing body weight ⁶⁹.

Finally, a useful negative. If raising GDF15 were the general mechanism of drug-induced weight loss, the glucagon-like peptide 1 (GLP-1) receptor agonists, the semaglutide family, ought to do it too. In a randomized trial of liraglutide for 26 weeks in 94 people with type 2 diabetes, weight fell and plasma GDF15 did not change at all, and the change in GDF15 was unrelated to the change in weight or in HbA1c, the share of hemoglobin carrying sugar, which reflects average blood glucose over the preceding weeks ⁷⁰. Whatever makes people nauseated on those drugs, it is not this hormone. One caution on reading that null: in a secondary analysis of a randomized trial in adults with obesity, circulating GDF15 rose transiently during the initial low-calorie-diet weight loss itself, so a drug trial that measures the hormone after weight has already fallen may be looking at the wrong moment ⁷¹.

Pregnancy: the clearest thing GDF15 does in people

Nausea and vomiting affect most pregnancies; the severe form, hyperemesis gravidarum, affects a small minority and can starve a woman. For most of the twentieth century the cause was unknown and the explanations offered were often about the woman's mind. The first hard genetic clue came in 2018, when a genome-wide association study of hyperemesis, run with data from the consumer genetics company 23andMe, found two loci above genome-wide significance and replicated them in an independent cohort: chromosome 19p13.11, which contains GDF15, and chromosome 4q12, which contains IGFBP7, the gene for insulin-like growth factor binding protein 7 ². Circulating evidence followed the same year: in 791 women in the Cambridge Baby Growth Study, serum GDF15 around week 15 was higher in those who vomited in the second trimester than in those with no nausea or vomiting, 11,581 against 10,593 pg/mL, and higher still in the eleven who took antiemetic drugs ⁷².

The decisive paper appeared online in December 2023. Using mass spectrometry to track a naturally labeled variant of the protein, the authors showed that the vast majority of GDF15 in a pregnant woman's plasma comes from the fetus and placenta, not from her. They then showed that the mother's own baseline matters more than the peak she reaches: women carrying genetic variants that lower their GDF15 outside pregnancy are at higher risk of hyperemesis, while women with beta-thalassemia, whose levels are chronically enormous, report very little nausea or vomiting in pregnancy. In mice, a prior long-acting dose blunted the response to a later acute dose, which is the mechanism the clinical pattern implies: desensitization ²¹. The absence of sickness in those women was reported in that same paper; the size of the levels it refers to comes from a separate cross-sectional study of 103 Sri Lankan patients, in which GDF15 averaged 24.2 times the healthy value in beta-thalassemia major ^21,25.

The 2026 multi-ancestry genome-wide study, of 10,974 cases and 461,461 controls across European, Asian, African and Latino ancestries, confirmed GDF15, IGFBP7, PGR, the progesterone receptor, and GFRAL, and added six further loci, among them TCF7L2, the best-known type 2 diabetes gene. Several of its findings bear directly on the mechanism. GDF15 and TCF7L2 are expressed mainly in extravillous trophoblasts, and the GDF15 signal acts in opposite directions depending on whether the variant is in the maternal or the fetal genome, which is what a hormone made by one person and sensed by another should look like. IGFBP7 and PGR, by contrast, were expressed mainly in the mother's spiral arteries, the vessels that feed the placenta, with effects confined to the maternal genome; and several of the loci were also associated with abnormal weight gain in pregnancy, pregnancy duration, birth weight and pre-eclampsia ⁴⁶.

Two practical consequences follow. The first is a biomarker. A longitudinal Danish study measured GDF15 repeatedly in early pregnancy: in normal pregnancies it rose steeply until week 7 and plateaued by week 9, pregnancies with female fetuses ran about 19 percent higher, the rise came a day later with male fetuses, and symptom scores peaked around week 8 as the hormone plateaued. In hyperemesis the hormone had already plateaued by the time of diagnosis, and although levels were higher early, the confidence intervals overlapped normal pregnancy by week 8. Early pregnancy loss showed lower levels and a slower rise ⁷³. That is a careful description of why a single measurement cannot diagnose the condition, and where a useful sampling window might be.

The second is prevention, and it is the most interesting unfinished business on this page. If low prepregnancy exposure is what sensitizes a woman, then raising her GDF15 before she conceives should protect her, and metformin raises GDF15. Two observational studies have now looked. A survey of 5,414 women recruited through a hyperemesis patient organization found that metformin use in the month before a first pregnancy was associated with a 71 percent lower risk of hyperemesis gravidarum, adjusted relative risk 0.29, 95 percent confidence interval 0.12 to 0.71, and that use before a second pregnancy was associated with an 82 percent lower risk of the combined outcome of severe nausea and vomiting of pregnancy or hyperemesis gravidarum, adjusted odds ratio 0.18, 0.06 to 0.59, after adjustment for the 86 percent recurrence rate seen in the same study ⁷⁴. The same questionnaire found associations in the opposite direction that deserve the same caution: before a first pregnancy, tobacco use carried an adjusted relative risk of 0.51, 0.30 to 0.86, and selective serotonin reuptake inhibitors 2.41, 1.33 to 4.38; before a second pregnancy, cannabis carried an adjusted odds ratio of 3.48, 1.80 to 6.75, and those same antidepressants 1.84, 1.12 to 3.04. All of it is self-reported, retrospective and drawn from a population recruited through a hyperemesis patient organization, so these are leads for trials rather than findings about cause ⁷⁴. A prospective cohort from the MotherToBaby programs compared 80 women exposed to metformin before conception with 4,411 unexposed: hyperemesis occurred in 1 of 80 against 97 of 4,411, adjusted relative risk 0.50, but the confidence interval ran from 0.07 to 3.39 and the study was not powered to detect a difference ⁷⁵. A 2026 review of both concludes that the mechanistic rationale is compelling, the clinical signals are encouraging, no randomized trial has yet shown preventive efficacy, and no practice recommendation can be made ³⁷.

One more role has appeared on the other side of the placenta. In mouse and pig models of fetal growth restriction, placental and fetal GDF15 are low and correlate with fetal weight; deleting GDF15 from the placenta alone causes growth restriction; and delivering GDF15 messenger RNA to the placenta in targeted lipid nanoparticles rescues it. The mechanism proposed is that placental GDF15 enters the fetal circulation and acts on the fetal liver through STAT3 to raise IGF2, a growth factor ²⁴. This is animal work and it points the opposite way from everything else on this page: for the fetus, more GDF15 is better.

As a blood test: strong prediction, weak causation

Within the populations where it has been measured, GDF15 is among the strongest single blood predictors of death and bleeding ^14,28,76, which is a smaller compliment than it sounds, because it predicts almost everything and specifies almost nothing. In the RE-LY trial of anticoagulation for atrial fibrillation, 8,474 patients were followed for a median of 1.9 years; those above 1,800 ng/L, compared with those below 1,200, had a hazard ratio for major bleeding of 1.76, 95 percent confidence interval 1.28 to 2.42, and for death of 1.72, 1.30 to 2.29, both after adjustment for clinical features and for other biomarkers including natriuretic peptide, troponin, interleukin-6, C-reactive protein and D-dimer. Annual rates of stroke also rose with GDF-15, and that association held after adjustment for clinical features, but unlike bleeding and death it did not survive further adjustment for the other biomarkers, which is why GDF-15 is used as a bleeding marker and not a stroke marker; adding it raised the c-index of the HAS-BLED score from 0.62 to 0.69 and of the ORBIT score from 0.68 to 0.71 ¹⁴. In the ARISTOTLE trial, with 14,798 patients and 1,272 deaths, a doubling of GDF-15 was the strongest biomarker predictor of death specifically from bleeding ⁷⁶. It is now one of three laboratory values in the ABC-AF bleeding score, whose updated version reached a C-index of 0.69 for major bleeding in 25,962 patients pooled from three randomized anticoagulation trials, better than the clinical scores ⁷⁷.

Outside atrial fibrillation the pattern holds. In 3,321 patients with coronary heart disease followed a median of 9.4 years, each natural-log unit of GDF-15 carried a hazard ratio of 1.49 for death and 1.29 for major cardiovascular events, and adding it to a Framingham-based model raised the C-index by 0.009, a real but small gain ¹⁹. In 310 patients having an aortic valve replaced through a catheter, GDF-15 predicted one-year mortality about as well as the EuroSCORE II surgical risk score, with areas under the curve of 0.686 and 0.711 ¹⁸.

The newest cardiology data are small and exploratory. In 84 consecutive patients having a tricuspid valve repaired through a catheter, baseline GDF-15 discriminated death at 12 months with an area under the curve of 0.823, 95 percent confidence interval 0.714 to 0.932, against 0.535, 0.362 to 0.709, for N-terminal pro-B-type natriuretic peptide, which did not discriminate at all. There were 10 deaths in the cohort, the authors call every analysis exploratory, and numbers built on 10 events will move ⁷⁸.

In mitochondrial disease it is diagnostic rather than merely prognostic. In 50 adults with primary mitochondrial myopathy it was raised in 86 percent, against 32 percent for lactate and 54 percent for creatine kinase, and the level tracked motor performance, lactate and the mutation load in muscle ²⁶. A review of thirteen candidate markers concluded that in large head-to-head comparisons GDF-15, followed by FGF-21, has the greatest value, while adding that neither is good enough to stand alone ⁷⁹.

Two caveats are structural rather than incidental. The first is analytical. A meta-analysis of genome-wide studies of GDF15 concentration across three assay platforms in up to 14,099 people found striking heterogeneity driven by a single common missense variant, rs1058587, which changes histidine to aspartate at position 202. The most plausible explanation is that the change sits in the epitope some antibodies recognize, so the assay, not the person, produces the difference ²⁸. Any study that compares GDF15 across people without accounting for that variant is measuring genotype as well as biology.

The second caveat is causal. The same study assessed 772 traits in 6,610 participants of FINRISK, a Finnish population cohort, and found GDF15 associated with all-cause mortality, cardiometabolic disease, respiratory disease, psychiatric disorders and inflammatory markers, which is the signature of a general stress readout; that it was a single national cohort limits how far those associations can be assumed to travel ²⁸. Mendelian randomization, which uses genetic variants as natural experiments, found evidence for a causal effect of GDF15 on high-density lipoprotein (HDL) cholesterol, the fraction usually called good cholesterol, but not on body mass index, and the reverse analysis found evidence that body mass index causes GDF15 rather than the other way round. The authors' conclusion is worth quoting in substance: human population data do not support moderately elevated GDF15 as a causal factor in cardiometabolic disease, but do support it as a biomarker of metabolic stress ²⁸.

• Serum or plasma GDF15 immunoassayAn antibody sandwich test on blood. Healthy volunteers average about 450 pg/mL; cardiology trials use 1,200 ng/L, the 90th percentile in healthy individuals, as the lower cut-point and 1,800 ng/L as the high-risk line ^13,14.Research, cardiovascular risk scores and mitochondrial disease work; it is not a routine test in most hospitalsA meta-analysis of genome-wide studies across three assay platforms in up to 14,099 people found significant heterogeneity between platforms, traced to the common missense variant rs1058587, so values are comparable within one platform but not reliably between them ²⁸.A common missense variant, rs1058587 (H202D), which swaps histidine for aspartate at position 202 of the protein, changes it where some antibodies bind; that produced a spurious genetic signal which differed by platform, so the difference is made by the assay rather than by the person, and a carrier's measured level is not comparable with a non-carrier's on every platform ²⁸.
• GDF15 as a marker of mitochondrial diseaseUsed to decide whether muscle weakness, exercise intolerance or a drooping eyelid comes from failing mitochondria, the structures that burn fuel inside cells ^26,79.Alongside FGF21 (fibroblast growth factor 21, another stress hormone released by failing tissue), lactate and creatine kinase in specialist neuromuscular clinicsIn 50 adults with primary mitochondrial myopathy it was raised in 86 percent, against 32 percent for lactate and 54 percent for creatine kinase, and it tracked motor performance and the mutation load in muscle ²⁶. In a deeply characterized m.3243A>G cohort of 102 patients it was one of 20 analytes that separated patients from controls ⁴⁷.It is raised by pregnancy, cancer, heart failure, kidney disease, infection, smoking and metformin, so a high value on its own means stress, not a mitochondrial diagnosis; a biomarker review judged it the best of thirteen candidates and still not good enough alone ⁷⁹.
• GDF-15 inside the ABC-AF bleeding score (age, biomarkers, clinical history, in atrial fibrillation)A score for people with atrial fibrillation on blood thinners that combines age, a history of bleeding, the type of anticoagulant taken (a direct oral anticoagulant or warfarin, added in version 2.0), and three laboratory values: GDF-15, hemoglobin and troponin T ⁷⁷.Estimating bleeding risk before and during anticoagulationIn 25,962 patients pooled from three randomized anticoagulation trials, with 1,321 major bleeds, the updated score reached a C-index of 0.69 for major bleeding and 0.72 for gastrointestinal bleeding, beating the clinical scores ⁷⁷.After adjustment for the other blood markers, GDF-15 still predicts bleeding and death but no longer stroke, so it informs one side of the anticoagulation decision only ¹⁴.
• GDF15 in early pregnancySerial measurement in the first trimester, when the hormone rises steeply from the placenta and fetus ⁷³.Research into nausea, vomiting and hyperemesis gravidarum; not yet a clinical testIn a longitudinal study, levels rose sharply until week 7 and plateaued by week 9, pregnancies carrying female fetuses ran about 19 percent higher, and symptom scores peaked around week 8 as the hormone plateaued ⁷³.The distributions overlap. Women with hyperemesis were higher only in the early window, with confidence intervals overlapping normal pregnancy by week 8, which is why a single value cannot diagnose the condition ⁷³.

The iron story, and a job GDF15 turned out not to have

In 2007 a group at the National Institutes of Health asked why people with thalassemia accumulate iron even when they are not transfused. The suspicion was that something released by the overworked marrow suppresses hepcidin, the liver hormone that restricts iron absorption. Profiling maturing red cell precursors, they found GDF15 rising sharply during maturation. Healthy volunteers averaged 450 plus or minus 50 pg/mL; patients with beta-thalassemia syndromes averaged 66,000 plus or minus 9,600, with individual values from 4,800 to 248,000. Serum from those patients suppressed hepcidin messenger RNA in cultured human liver cells, and removing GDF15 from the serum reversed the suppression ¹³. For several years GDF15 was taught as the erythroid regulator of iron ^13,40.

In 2014 a different molecule took the job. Erythroferrone, produced by red cell precursors in response to erythropoietin, was identified as the mediator of hepcidin suppression after blood loss: mice lacking it fail to suppress hepcidin after hemorrhage and recover slowly, and it is greatly increased in a mouse model of thalassemia intermedia, where it contributes to the hepcidin suppression and the iron overload ³⁹. A review the following year puts the current position carefully: erythroferrone is the mediator of hepcidin suppression during stress erythropoiesis, and erythroferrone together with other candidates, including GDF15, may contribute in ineffective erythropoiesis, the specific situation of thalassemia where precursors are produced and destroyed in the marrow ⁴⁰.

This is a reversal worth naming because of how it happened. The 2007 observation was real, reproducible and large: GDF15 in thalassemia is genuinely among the highest concentrations of any human hormone, and thalassemia serum genuinely suppresses hepcidin. What the field got wrong was the inference from correlation plus plausibility to physiological role. The molecule that does the job in normal stress erythropoiesis is a different one, and GDF15's enormous levels in thalassemia are now better understood through their other consequence: the low body weight and poor growth of those patients, and their striking freedom from pregnancy sickness ^21,25.

Aging and longevity, at their true strength

GDF15 is one of the most reliable protein markers of human age. Across a systematic review of 36 proteomic studies of aging, 1,128 proteins were reported to change with age by two or more analyses, and GDF15 is among the 32 reported by five or more ⁵³. In a cross-sectional study of 228 people aged 18 to 83 it rose with age, was lower in habitually active cyclists than in healthy controls from the general population, and was highest in people immobilized by lower-limb disease, correlating with quadriceps torque and with leukocyte counts ⁵². Reading that honestly: in people, GDF15 marks inactivity, inflammation and weak muscle, and rises with age.

The longevity claim is narrower than it is usually reported. Female mice carrying a human GDF15 transgene had significantly longer mean and median lifespans in two independent transgenic lines, with the difference larger on a high-fat diet than a low-fat one; they were leaner, had lower IGF-1, insulin and glucose, better insulin sensitivity, higher oxygen use and reduced signaling through the insulin-like growth factor 1 and mTOR pathways associated with longevity, mTOR being the mechanistic target of rapamycin, the master switch that tells a cell to grow ⁸⁰. That is a real mouse lifespan experiment, in females, on a background where the animals were protected from diet-induced obesity. It is not evidence that GDF15 extends human life, and the human epidemiology points the other way: high GDF15 is associated with higher, not lower, mortality in the clinical cohorts that have measured it ^14,19,28.

Those two observations are less contradictory than they look. A chronically elevated hormone of cellular distress in an old person is reporting damage, and the damage is what kills them. A transgene that raises the same hormone in a healthy young mouse keeps it thin. The claim that would unify them, that GDF15 signaling is beneficial when it moderates intake and harmful only as a messenger of underlying disease, has not been tested in people, and the gerontologists arguing against long-term GDF15 blockade are making the inverse of the same untested claim ⁶³.

One more human observation belongs here, as a preprint and therefore as a hypothesis rather than a finding. A 2026 study, not yet peer reviewed, reports that circulating GDF15 is about fivefold higher in people with alcohol dependence, that a rare loss-of-function variant in GFRAL is associated with drinking about 2.6 additional UK units, roughly 21 grams of ethanol, per week in a large biobank, and that recombinant GDF15 reduces alcohol drinking in mice. The authors propose that GDF15 is an endocrine brake on alcohol intake ⁸¹. Its appeal is that it fits the rest of the page: a molecule released by poisoned tissue that tells the brainstem to stop taking the poison in.

The obesity drug that did not work, and the knockouts that made blocking it look safe

If a hormone reliably makes animals eat less and lose weight, the obvious commercial move is to give it to people with obesity. Eli Lilly built LY3463251, a long-acting GDF15 analogue that activates GFRAL. In rodents and in obese nonhuman primates it reduced food intake and body weight with no signs of malaise or vomiting. In healthy human volunteers, single injections produced dose-dependent nausea and vomiting in some people, and in a 12-week dose-ranging study in overweight and obese participants it significantly reduced food intake and appetite scores but produced only modest weight loss, which was not explained by nausea. The authors' own conclusion was that the GFRAL system can modulate human energy balance but that the weight effect is surprisingly small, suggesting real difficulty in using this pathway for weight loss ¹⁷. It is an unusually candid negative result from a sponsor, and it is the main reason GDF15 agonism is no longer a leading obesity strategy while GLP-1 agonism is.

Pushing in the other direction, the safety question for the cachexia and oncology drugs is what happens to a person who has no GDF15 at all. A 2024 study answered it directly. Among 75,018 participants with whole-exome or whole-genome sequencing in the Pakistan Genomic Resource, the investigators identified and then recalled carriers of stop-gain, frameshift and fully inactivating missense variants, including C211G: 8 homozygous knockouts and 227 heterozygous carriers. The knockouts ranged in age from 31 to 75, were fertile, had multiple children, and showed no consistent phenotype, including no metabolic abnormality. The authors conclude that GDF15 is not required for fertility, healthy pregnancy, fetal development or survival into adulthood, and that this supports the safety of drugs that block it ²³. The work was done by Novartis with collaborators from AstraZeneca, Astellas and Regeneron, all of which have an interest in the answer; the design, a recall-by-genotype study in a consanguineous population where homozygotes are findable, is nonetheless the strongest available test.

Note the small tension between these two papers. The Novartis authors summarize the pharmacology as GDF15 agonism causing nausea in humans without accompanying weight loss ²³, while the Lilly trial reported a modest but significant weight reduction ¹⁷. Both are describing the same disappointing effect size from different positions, and the honest statement is that activating GFRAL in people makes them feel sick reliably and makes them thinner only slightly.

The current drug landscape, as a 2026 medicinal-chemistry review describes it, runs in both directions at once: agonists of GFRAL for appetite suppression, antagonists and antibodies that clear circulating GDF15 for cachexia, nausea and immunotherapy resistance, in biologics and small molecules, with patents and trials on both sides of the same axis ⁸².

What was believed and is not, and what is still unknown

Five claims about GDF15 have been overturned or substantially revised, and they are worth listing because each was reasonable when made. It was believed that GDF15 acts on the hypothalamus through the type II transforming growth factor beta receptor; the receptor is GFRAL, with RET, and it is in the hindbrain only ^6,8,38. It was believed that GDF15 is the erythroid signal that suppresses hepcidin; that role belongs to erythroferrone, with GDF15 at most contributing in ineffective erythropoiesis ^13,39,40. It was believed that metformin's weight-lowering effect runs through GDF15 and GFRAL; one group showed that and a second could not reproduce the requirement, and the question is open ^30,42,43. And it was believed, implicitly, that GDF15 reduces food intake by acting on appetite; the evidence from species that can vomit says it works by causing nausea ^32,36. And it was believed that GDF15 lowers body weight purely by lowering food intake; in mice it also defends energy expenditure during caloric restriction, so that account is incomplete rather than wrong ^8,44.

A sixth reversal, methodological, underlies several of the others. Recombinant GDF15 made in mammalian cells carries trace transforming growth factor beta, which is biologically active at femtomolar concentrations; a 2017 paper demonstrated that effects attributed to GDF15 in their own hands were caused by that contaminant, and warned that previously published conclusions may not hold ⁴¹. Any pre-2017 experiment using purified recombinant GDF15 without an appropriate control has to be read with that in mind.

What is genuinely established: the gene and the protein's structure ^1,3,5; the identity and signaling partner of the receptor, and that the hindbrain is where it has been found ^6,7,8,20,34; that GDF15 causes vomiting in a species that vomits ³²; that most GDF15 in maternal blood is fetal, and that low prepregnancy exposure predicts severe pregnancy sickness ²¹; that neutralizing it adds weight in cancer cachexia over 12 weeks ¹⁵; that it is a powerful and nonspecific prognostic marker ^14,28; and that people born without it appear to be well ²³.

What is not known. Whether blocking GDF15 helps or harms over months to years, in cachexia or anywhere else, because no trial has measured survival ^15,63. Whether metformin's benefits depend on it ⁴³. Whether preconception metformin prevents hyperemesis gravidarum, which two observational studies suggest and no randomized trial has tested ^37,74,75. Whether GDF15 has any role in cyclic vomiting syndrome or cannabinoid hyperemesis, both of which have the right phenotype and neither of which has published human GDF15 data; a 2025 review of emesis lists cyclic vomiting syndrome, alongside chemotherapy-induced emesis and hyperemesis gravidarum, as a condition where the measurements have not been made ³⁶, while cannabinoid hyperemesis appears as a candidate pathway only in the cyclic-vomiting literature, where GDF-15 is one of several unconfirmed mechanisms named beside the receptor for substance P, the first cannabinoid receptor, calcitonin gene-related peptide and mitochondrial dysfunction ⁸³. Whether GFRAL exists anywhere outside the hindbrain, which later reports raise and no study has settled ⁹. Whether GFRAL neurons are a nutrient sensor rather than a poison alarm, which one 2026 preprint argues and peer review has not yet tested ⁵⁴. Whether the single-arm oncology responses to GDF-15 blockade survive a randomized comparison ⁶⁴. And what, if anything, GDF15 does inside the tissues that store it in latent form, which is most of the protein the body makes and none of what the blood test sees ^29,33.

Finally, a caution about reading any GDF15 number in the literature. The common H202D variant distorts some immunoassays, so cross-person comparisons may be partly comparing genotypes ²⁸; and the field's own population genetics concludes that moderately raised GDF15 is a marker of metabolic stress and not, on current evidence, a cause of cardiometabolic disease ²⁸. It is a thermometer that occasionally also sets fires, and telling those two roles apart, condition by condition, is most of the work still to be done ^66,82.

Where it connects

Sources

83 sources, numbered as they are cited. Every one was checked against PubMed or its publisher before it was cited here; the note under each says what it shows and what it does not.

1. 1
   Lawton LN, Bonaldo MF, Jelenc PC, et al. Identification of a novel member of the TGF-beta superfamily highly expressed in human placenta.Gene · 1997 · 203(1):17-26doi:10.1016/s0378-1119(97)00485-x · PMID 9426002

   Names it PTGFB; 1.2-kb mRNA, putative pre-pro protein of 295 amino acids and mature protein of 112; single-copy gene of two exons (309 and 891 bp) separated by a 2.9-kb intron; mapped to 19p12-13.1 by fluorescence in situ hybridization; a second 1.9-kb transcript predominates in adult skeletal muscle.

2. 2
   Fejzo MS, Sazonova OV, Sathirapongsasuti JF, et al. Placenta and appetite genes GDF15 and IGFBP7 are associated with hyperemesis gravidarum.Nat Commun · 2018 · 9(1):1178doi:10.1038/s41467-018-03258-0 · PMID 29563502

   Genome-wide association study of hyperemesis and of nausea and vomiting severity, run with 23andMe data: chromosome 19p13.11 (GDF15) and 4q12 (IGFBP7) reached genome-wide significance and replicated. The authors state that proving a causal role requires further study.

3. 3
   Yokoyama-Kobayashi M, Saeki M, Sekine S, Kato S. Human cDNA encoding a novel TGF-beta superfamily protein highly expressed in placenta.J Biochem · 1997 · 122(3):622-6doi:10.1093/oxfordjournals.jbchem.a021798 · PMID 9348093

   Independent 1997 cloning by a secretory-signal screen; reports a 308-amino-acid protein with seven conserved C-terminal cysteines preceded by a basic processing site, highly expressed in placenta. The 295 versus 308 difference with the Gene paper is a start-codon question, not a difference in the mature protein.

4. 4
   Paralkar VM, Vail AL, Grasser WA, et al. Cloning and characterization of a novel member of the transforming growth factor-beta/bone morphogenetic protein family.J Biol Chem · 1998 · 273(22):13760-7doi:10.1074/jbc.273.22.13760 · PMID 9593718

   Names it prostate-derived factor; two cysteines in addition to the seven conserved; high expression in placenta and prostate; androgen-regulated in rat prostate; subcutaneous implantation of recombinant protein induced cartilage and early endochondral bone. Work done at Pfizer.

5. 5
   Bootcov MR, Bauskin AR, Valenzuela SM, et al. MIC-1, a novel macrophage inhibitory cytokine, is a divergent member of the TGF-beta superfamily.Proc Natl Acad Sci U S A · 1997 · 94(21):11514-9doi:10.1073/pnas.94.21.11514 · PMID 9326641

   The founding paper. Subtraction cloning from activated macrophages; a divergent superfamily member with no strong homology to existing families; induced by IL-1beta, TNF-alpha, IL-2, M-CSF and TGF-beta but not IFN-gamma or lipopolysaccharide; secreted from CHO cells as a 25 kDa cysteine-rich dimer after propeptide cleavage; recombinant protein inhibited macrophage TNF-alpha production. Nothing about appetite.

6. 6
   Mullican SE, Lin-Schmidt X, Chin CN, et al. GFRAL is the receptor for GDF15 and the ligand promotes weight loss in mice and nonhuman primates.Nat Med · 2017 · 23(10):1150-1157doi:10.1038/nm.4392 · PMID 28846097

   Janssen. Weight loss in high-fat-fed mice and spontaneously obese nonhuman primates; high-affinity binding to GFRAL; Gfral expression in area postrema and solitary nucleus neurons in mice and humans; receptor deletion abolished the effect; signaling requires the co-receptor RET.

7. 7
   Emmerson PJ, Wang F, Du Y, et al. The metabolic effects of GDF15 are mediated by the orphan receptor GFRAL.Nat Med · 2017 · 23(10):1215-1219doi:10.1038/nm.4393 · PMID 28846098

   Eli Lilly, published the same day. GFRAL binds GDF15 with high affinity; Gfral-deficient mice were refractory to effects on body weight, food intake and glucose; a blocking monoclonal antibody prevented the effects in rats; Gfral mRNA highly expressed in the area postrema of mouse, rat and monkey.

8. 8
   Yang L, Chang CC, Sun Z, et al. GFRAL is the receptor for GDF15 and is required for the anti-obesity effects of the ligand.Nat Med · 2017 · 23(10):1158-1166doi:10.1038/nm.4394 · PMID 28846099

   Novo Nordisk, the third same-day paper. Establishes that RET is required for signaling, that GFRAL expression is limited to hindbrain neurons and absent from peripheral tissues, and that GDF15 did not raise energy expenditure in obese mice, from which the authors concluded that the anti-obesity action is likely driven primarily by reduced food intake; that last conclusion was qualified by mouse work in 2023.

9. 9
   Di Santo A, Tarchi L, Villa G, et al. GDF15 Analogues Acting as GFRAL Ligands.ChemMedChem · 2025 · 20(9):e202400961doi:10.1002/cmdc.202400961 · PMID 39907315

   Review of GDF15-derived agonists and antagonists. States that recent studies suggest a broader GFRAL distribution than the hindbrain alone, which may explain distinct roles of GDF15. A review, and the broader-distribution claim is not itself primary data.

10. 10
    Hromas R, Hufford M, Sutton J, Xu D, Li Y, Lu L. PLAB, a novel placental bone morphogenetic protein.Biochim Biophys Acta · 1997 · 1354(1):40-4doi:10.1016/s0167-4781(97)00122-x · PMID 9375789

    Names it PLAB; most closely related to BMP-8/OP-2; highly expressed in placenta, low elsewhere; inhibits proliferation of primitive hematopoietic progenitors. A fourth independent naming in the same year.

11. 11
    Baek SJ, Kim KS, Nixon JB, Wilson LC, Eling TE. Cyclooxygenase inhibitors regulate the expression of a TGF-beta superfamily member that has proapoptotic and antitumorigenic activities.Mol Pharmacol · 2001 · 59(4):901-8PMID 11259636

    NSAID-activated gene 1, found by subtraction cloning in drug-treated HCT-116 colon cancer cells; states explicitly that the sequence is identical to a protein that already has five different names; transfected cells showed more apoptosis and less colony formation, and derived tumors grew less in nude mice. No DOI is registered for this article in PubMed.

12. 12
    Bottner M, Laaff M, Schechinger B, Rappold G, Unsicker K, Suter-Crazzolara C. Characterization of the rat, mouse, and human genes of growth/differentiation factor-15/macrophage inhibiting cytokine-1 (GDF-15/MIC-1).Gene · 1999 · 237(1):105-11doi:10.1016/s0378-1119(99)00309-1 · PMID 10524241

    Genomic library screening in three species: rat, mouse and human genes each of two exons with a single intron interrupting the coding sequence at identical positions within the prepro-domain; the seven conserved carboxy-terminal cysteines; abundant messenger RNA in many embryonic and adult tissues; promoter analysis in the rat. The orthologs show the lowest sequence conservation of any member of the TGF-beta superfamily. The paper that fixed the name GDF-15.

13. 13
    Tanno T, Bhanu NV, Oneal PA, et al. High levels of GDF15 in thalassemia suppress expression of the iron regulatory protein hepcidin.Nat Med · 2007 · 13(9):1096-101doi:10.1038/nm1629 · PMID 17721544

    GDF15 rises during erythroblast maturation; healthy volunteers averaged 450 plus or minus 50 pg/mL against 66,000 plus or minus 9,600 in beta-thalassemia (range 4,800 to 248,000); thalassemia serum suppressed hepcidin mRNA in primary human hepatocytes and depleting GDF15 reversed the suppression. The source of the healthy reference value used throughout this page, and the origin of a role later reassigned.

14. 14
    Hijazi Z, Oldgren J, Andersson U, et al. Growth-differentiation factor 15 and risk of major bleeding in atrial fibrillation: Insights from the Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) trial.Am Heart J · 2017 · 190:94-103doi:10.1016/j.ahj.2017.06.001 · PMID 28760218

    8,474 patients, median 1.9 years. Cut-points of 1,200 ng/L (the 90th percentile in healthy individuals) and 1,800 ng/L; highest versus lowest group, major bleeding hazard 1.76 (1.28-2.42) and death 1.72 (1.30-2.29) after adjustment for other biomarkers; annual stroke rates also rose with GDF-15, but that association did not survive adjustment for the other biomarkers; improved the c-index of HAS-BLED from 0.62 to 0.69 and of ORBIT from 0.68 to 0.71. A biomarker substudy of a randomized trial.

15. 15
    Groarke JD, Crawford J, Collins SM, et al. Ponsegromab for the Treatment of Cancer Cachexia.N Engl J Med · 2024 · 391(24):2291-2303doi:10.1056/NEJMoa2409515 · PMID 39282907

    Randomized, double-blind, 12-week phase 2 trial, 187 patients with cancer cachexia and GDF-15 at or above 1500 pg/mL; median weight differences against placebo of 1.22, 1.92 and 2.81 kg at 100, 200 and 400 mg; appetite, symptoms and activity improved at 400 mg; adverse events 70 percent on drug, 80 percent on placebo. Funded by Pfizer; NCT05546476. Weight, not survival, and 12 weeks only.

16. 16
    Melero I, de Miguel Luken M, de Velasco G, et al. Neutralizing GDF-15 can overcome anti-PD-1 and anti-PD-L1 resistance in solid tumours.Nature · 2025 · 637(8048):1218-1227 (online 11 Dec 2024)doi:10.1038/s41586-024-08305-z · PMID 39663448

    First-in-human GDFATHER-1/2a trial (NCT04725474) of visugromab plus nivolumab in tumors refractory to anti-PD-1 or anti-PD-L1; durable deep responses in some non-squamous lung and urothelial cancers, the two types flagged by an in silico screen of about 10,000 tumors in The Cancer Genome Atlas; more T cell infiltration and interferon-gamma signaling after treatment. Single-arm, sponsored by CatalYm, whose employees are among the authors.

17. 17
    Benichou O, Coskun T, Gonciarz MD, et al. Discovery, development, and clinical proof of mechanism of LY3463251, a long-acting GDF15 receptor agonist.Cell Metab · 2023 · 35(2):274-286.e10doi:10.1016/j.cmet.2022.12.011 · PMID 36630958

    A GFRAL agonist that reduced food intake and weight in rodents and obese nonhuman primates without malaise; in humans, dose-dependent nausea and emesis in some participants, and over 12 weeks in overweight and obese adults, significant reductions in food intake and appetite with only modest weight loss (NCT03764774). The sponsor concludes the pathway is hard to use for weight loss.

18. 18
    Sinning JM, Wollert KC, Sedaghat A, et al. Risk scores and biomarkers for the prediction of 1-year outcome after transcatheter aortic valve replacement.Am Heart J · 2015 · 170(4):821-9doi:10.1016/j.ahj.2015.07.003 · PMID 26386807

    310 consecutive patients; of four surgical risk scores and five biomarkers, EuroSCORE II and GDF-15 were the strongest predictors of one-year death (areas under the curve 0.711 and 0.686), and GDF-15 reclassified patients when added to the scores. Observational.

19. 19
    Lyu L, Xv C, Xu J, et al. Growth differentiation factor-15 predicts all-cause death and major adverse cardiovascular events in patients with coronary heart disease: a prospective cohort study.J Thromb Thrombolysis · 2024 · 57(7):1109-1121doi:10.1007/s11239-024-03019-5 · PMID 39068629

    Single-center cohort of 3,321 patients, median 9.4 years, 759 deaths and 1,291 major events. Hazard 1.49 (1.36-1.64) per natural-log unit for death and 1.29 (1.20-1.38) for major events; tertiles below 1,370, 1,370 to 2,556 and above 2,556 ng/L; adding GDF-15 to a Framingham-based model raised the C-index by 0.009. Observational and single-center.

20. 20
    Hsu JY, Crawley S, Chen M, et al. Non-homeostatic body weight regulation through a brainstem-restricted receptor for GDF15.Nature · 2017 · 550(7675):255-259doi:10.1038/nature24042 · PMID 28953886

    NGM Biopharmaceuticals, a month after the three Nature Medicine papers. Gfral-knockout mice were hyperphagic under stress and resistant to chemotherapy-induced anorexia and weight loss; traces the circuit from area postrema and solitary nucleus to the parabrachial nucleus and central amygdala.

21. 21
    Fejzo M, Rocha N, Cimino I, et al. GDF15 linked to maternal risk of nausea and vomiting during pregnancy.Nature · 2024 · 625(7996):760-767 (online 13 Dec 2023)doi:10.1038/s41586-023-06921-9 · PMID 38092039

    Mass spectrometry of a naturally labeled GDF15 variant showed that the vast majority of maternal plasma GDF15 comes from the feto-placental unit; rare and common variants lowering non-pregnant GDF15 raised hyperemesis risk; women with beta-thalassemia, chronically very high, reported very little nausea or vomiting; in mice the acute food-intake response was blunted by prior exposure, indicating desensitization. The pivotal paper for the page.

22. 22
    Temel JS, Abernethy AP, Currow DC, et al. Anamorelin in patients with non-small-cell lung cancer and cachexia (ROMANA 1 and ROMANA 2): results from two randomised, double-blind, phase 3 trials.Lancet Oncol · 2016 · 17(4):519-531doi:10.1016/S1470-2045(15)00558-6 · PMID 26906526

    Two randomized placebo-controlled phase 3 trials, 484 and 495 patients with stage III or IV non-small-cell lung cancer and cachexia. Over 12 weeks median lean body mass rose 0.99 kg (95 percent CI 0.61 to 1.36) on anamorelin against -0.47 kg (-1.00 to 0.21) on placebo in ROMANA 1, and 0.65 kg (0.38 to 0.91) against -0.98 kg (-1.49 to -0.41) in ROMANA 2, both p<0.0001; handgrip strength did not differ in either trial. Funded by Helsinn Therapeutics. The pre-GDF15 benchmark for drug effects in cancer cachexia, and lean body mass is not survival.

23. 23
    Gurtan AM, Khalid S, Koch C, et al. Identification and characterization of human GDF15 knockouts.Nat Metab · 2024 · 6(10):1913-1921doi:10.1038/s42255-024-01135-3 · PMID 39327531

    From 75,018 sequenced participants in the Pakistan Genomic Resource, 8 homozygous and 227 heterozygous carriers of loss-of-function GDF15 variants, including C211G; the knockouts, aged 31 to 75, were fertile with multiple children and showed no consistent phenotype. Read by the authors as supporting the safety of blocking GDF15. Performed by Novartis with AstraZeneca, Astellas and Regeneron collaborators, all with an interest in the answer.

24. 24
    Liu XP, Zhang X, Cai CX, et al. Deficiency of placental GDF15 causes fetal growth restriction by inhibiting fetal hepatic STAT3-IGF2 axis.PLoS Biol · 2026 · 24(10):e3004061doi:10.1371/journal.pbio.3004061 · PMID 42873832

    In mouse and pig models of fetal growth restriction, placental and fetal GDF15 are reduced and correlate with fetal weight; placenta-specific deletion causes growth restriction, and placenta-targeted Gdf15 mRNA in lipid nanoparticles rescues it, through fetal hepatic STAT3 and IGF2. Animal work; it argues that for the fetus more GDF15 is better, the opposite of the maternal story.

25. 25
    Karusheva Y, Petry CJ, Yasara N, et al. Association of GDF15 levels with body mass index and endocrine status in beta-thalassaemia.Clin Endocrinol (Oxf) · 2023 · 99(2):182-189doi:10.1111/cen.14897 · PMID 36806122

    Cross-sectional study of 103 patients: GDF15 24.2-fold above healthy controls in beta-thalassemia major; everyone above 24,000 pg/mL had a BMI below 20 kg/m2; IGF-1 correlated inversely after adjustment. Cross-sectional, so the direction of the BMI relationship is inferred from other evidence.

26. 26
    Martin-Jimenez P, Bermejo-Guerrero L, Navarro-Riquelme M, et al. Comprehensive analysis of GDF15 as a biomarker in primary mitochondrial myopathies.Mol Genet Metab · 2025 · 144(3):109023doi:10.1016/j.ymgme.2025.109023 · PMID 39854975

    Fifty adults with primary mitochondrial myopathy: GDF15 raised in 86 percent against 32 percent for lactate and 54 percent for creatine kinase, correlating with motor performance, lactate and mtDNA mutation load in muscle; reduced creatinine emerged as a second marker. Observational, single center, no control group of other myopathies.

27. 27
    Breen DM, Kim H, Bennett D, et al. GDF-15 Neutralization Alleviates Platinum-Based Chemotherapy-Induced Emesis, Anorexia, and Weight Loss in Mice and Nonhuman Primates.Cell Metab · 2020 · 32(6):938-950.e6doi:10.1016/j.cmet.2020.10.023 · PMID 33207247

    Pfizer. Circulating GDF-15 higher in people on platinum chemotherapy and positively associated with weight loss in colorectal cancer (NCT00609622); agents with high clinical emetic scores raised GDF-15 in mice; knockout mice were protected; a neutralizing antibody improved survival in mice and reduced anorexia and emesis in nonhuman primates.

28. 28
    Lemmela S, Wigmore EM, Benner C, et al. Integrated analyses of growth differentiation factor-15 concentration and cardiometabolic diseases in humans.Elife · 2022 · 11:e76272doi:10.7554/eLife.76272 · PMID 35916366

    Up to 14,099 individuals across three assay platforms, and 772 traits in 6,610 participants of the Finnish population cohort FINRISK. Finds heterogeneity across three assay platforms driven by the common missense variant rs1058587 (p.H202D), most plausibly an epitope-binding artifact; Mendelian randomization supports a causal effect on HDL cholesterol but not on BMI, with reverse causation from BMI to GDF15. Concludes GDF15 is a biomarker of metabolic stress, not a causal factor in cardiometabolic disease. The key caution for reading any GDF15 measurement.

29. 29
    Bauskin AR, Jiang L, Luo XW, Wu L, Brown DA, Breit SN. The TGF-beta superfamily cytokine MIC-1/GDF15: secretory mechanisms facilitate creation of latent stromal stores.J Interferon Cytokine Res · 2010 · 30(6):389-97doi:10.1089/jir.2009.0052 · PMID 20187768

    Induction by proinflammatory cytokines and oxidized LDL, abundance in plaque and tumor macrophages, status as a major p53 target gene; unprocessed precursor secreted rapidly and mature protein slowly, with the C-terminal 47 residues of the propeptide responsible. Cell-line work in U937 monocytoid cells.

30. 30
    Day EA, Ford RJ, Smith BK, et al. Metformin-induced increases in GDF15 are important for suppressing appetite and promoting weight loss.Nat Metab · 2019 · 1(12):1202-1208doi:10.1038/s42255-019-0146-4 · PMID 32694673

    Metformin induced GDF15 in mouse hepatocytes through ATF4 and CHOP; effects on food intake, body mass, fasting insulin and glucose tolerance were eliminated in Gdf15-null mice; rising serum GDF15 was associated with weight loss in patients with type 2 diabetes. Published within weeks of the Nature paper, with Sanofi coauthors.

31. 31
    Shah FA, Bahudhanapati H, Jiang M, et al. Lung Epithelium Releases Growth Differentiation Factor 15 in Response to Pathogen-mediated Injury.Am J Respir Cell Mol Biol · 2024 · 70(5):379-391doi:10.1165/rcmb.2023-0429OC · PMID 38301257

    Human biospecimens plus mouse models. In critically ill patients, plasma GDF15 correlated with lower respiratory tract levels and was higher in nonsurvivors; SARS-CoV-2 induced GDF15 in human lung epithelium. In mice, epithelial-specific deletion attenuated airspace and plasma release after Pseudomonas aeruginosa type II secretion system exoproducts, and airspace reconstitution raised erythrocyte counts. The human part is observational and prognostic; the mechanism is mouse.

32. 32
    Borner T, Shaulson ED, Ghidewon MY, et al. GDF15 Induces Anorexia through Nausea and Emesis.Cell Metab · 2020 · 31(2):351-362.e5doi:10.1016/j.cmet.2019.12.004 · PMID 31928886

    GDF15 caused vomiting in musk shrews and nausea-like behaviors, including pica, in mice and in lean and obese rats; also contributed to chemotherapy-induced malaise in mice. The paper that reframed the anorexia as nausea. Animal work throughout.

33. 33
    Bauskin AR, Brown DA, Junankar S, et al. The propeptide mediates formation of stromal stores of PROMIC-1: role in determining prostate cancer outcome.Cancer Res · 2005 · 65(6):2330-6doi:10.1158/0008-5472.CAN-04-3827 · PMID 15781647

    Only unprocessed proMIC-1 binds the extracellular matrix, via the C-terminal 89 amino acids of the propeptide; removing the propeptide raised serum levels about twentyfold in xenografts; stromal levels in prostate cancer tissue predicted relapse after prostatectomy. Tissue cores and xenografts, not a randomized study.

34. 34
    Li J, Shang G, Chen YJ, et al. Cryo-EM analyses reveal the common mechanism and diversification in the activation of RET by different ligands.Elife · 2019 · 8:e47650doi:10.7554/eLife.47650 · PMID 31535977

    Structures of the extracellular GDF15/GFRAL/RET ternary complex alongside three related complexes; all four ligand and co-receptor pairs induce the same RET dimer, positioned to allow the two kinase domains to cross-phosphorylate. Structural biology, no physiology.

35. 35
    Tsai VW, Manandhar R, Jorgensen SB, et al. The anorectic actions of the TGFbeta cytokine MIC-1/GDF15 require an intact brainstem area postrema and nucleus of the solitary tract.PLoS One · 2014 · 9(6):e100370doi:10.1371/journal.pone.0100370 · PMID 24971956

    Intraperitoneal GDF15 induced Fos in the area postrema and medial nucleus of the solitary tract, in neurons not staining for tyrosine hydroxylase; combined ablation of both nuclei completely reversed the anorectic effect. Localized the site three years before the receptor was found.

36. 36
    Borner T, Pataro AM, De Jonghe BC. Central mechanisms of emesis: A role for GDF15.Neurogastroenterol Motil · 2025 · 37(3):e14886 (online 6 Aug 2024)doi:10.1111/nmo.14886 · PMID 39108013

    Review arguing that GDF15 causes anorexia through nausea, emesis and food aversion rather than through appetite regulation, and listing chemotherapy-induced emesis, hyperemesis gravidarum and cyclic vomiting syndrome as candidate conditions; cannabinoid hyperemesis is not among them. A review, and the cyclic vomiting link is a hypothesis, not data.

37. 37
    Deruelle P, Arcieri F, Martineau R. Metformin for the prevention of hyperemesis gravidarum: From a GDF15-based rationale to early clinical evidence.Gynecol Obstet Fertil Senol · 2026 · online 6 Jun 2026doi:10.1016/j.gofs.2026.06.002 · PMID 42250761

    French-language review of the desensitization rationale and of the two observational datasets; concludes that the mechanism is compelling and the signals encouraging but that no randomized trial has demonstrated preventive efficacy and no recommendation can be made.

38. 38
    Johnen H, Lin S, Kuffner T, et al. Tumor-induced anorexia and weight loss are mediated by the TGF-beta superfamily cytokine MIC-1.Nat Med · 2007 · 13(11):1333-40doi:10.1038/nm1677 · PMID 17982462

    Serum MIC-1 tracked weight loss in advanced prostate cancer; xenograft mice wasted and were rescued by antibody; systemic and transgenic MIC-1 caused hypophagia. Its proposed mechanism, hypothalamic TGF-beta receptor II with ERK, STAT3, NPY and POMC, was superseded in 2017.

39. 39
    Kautz L, Jung G, Valore EV, Rivella S, Nemeth E, Ganz T. Identification of erythroferrone as an erythroid regulator of iron metabolism.Nat Genet · 2014 · 46(7):678-84doi:10.1038/ng.2996 · PMID 24880340

    Erythroferrone, produced by erythroblasts in response to erythropoietin, mediates hepcidin suppression during stress erythropoiesis; knockout mice fail to suppress hepcidin after hemorrhage, and erythroferrone is greatly increased in a mouse model of thalassemia intermedia. The molecule that took the job GDF15 had been given.

40. 40
    Kim A, Nemeth E. New insights into iron regulation and erythropoiesis.Curr Opin Hematol · 2015 · 22(3):199-205doi:10.1097/MOH.0000000000000132 · PMID 25710710

    Review stating the settled position: erythroferrone is the mediator of hepcidin suppression in stress erythropoiesis, with erythroferrone and additional candidates including TWSG1 and GDF15 possibly contributing in ineffective erythropoiesis. A review, and deliberately hedged on GDF15.

41. 41
    Olsen OE, Skjaervik A, Stordal BF, Sundan A, Holien T. TGF-beta contamination of purified recombinant GDF15.PLoS One · 2017 · 12(11):e0187349doi:10.1371/journal.pone.0187349 · PMID 29161287

    Recombinant GDF15 batches from mammalian expression systems carried TGF-beta, active at femtomolar concentrations; effects the authors first attributed to GDF15 were caused by the contaminant. States that some already published conclusions about GDF15 may not be supported.

42. 42
    Coll AP, Chen M, Taskar P, et al. GDF15 mediates the effects of metformin on body weight and energy balance.Nature · 2020 · 578(7795):444-448doi:10.1038/s41586-019-1911-y · PMID 31875646

    Two independent randomized controlled trials showing metformin raises circulating GDF15 in people; in mice the rise came mainly from distal intestine and kidney; metformin prevented high-fat-diet weight gain in wild-type but not Gdf15- or Gfral-null mice, and a GFRAL-antagonist antibody reversed it in obese mice; glucose lowering was preserved without GDF15. The human part is a biomarker measurement, not a demonstration that the pathway is required in people.

43. 43
    Klein AB, Nicolaisen TS, Johann K, et al. The GDF15-GFRAL pathway is dispensable for the effects of metformin on energy balance.Cell Rep · 2022 · 40(8):111258doi:10.1016/j.celrep.2022.111258 · PMID 36001956

    Confirms that metformin raises GDF15 in mice and humans but fails to confirm that the pathway is necessary: metformin still lowered body weight in GDF15-knockout and GFRAL-knockout mice. The direct contradiction of the 2019-2020 claim; both sides used knockout animals.

44. 44
    Wang D, Townsend LK, DesOrmeaux GJ, et al. GDF15 promotes weight loss by enhancing energy expenditure in muscle.Nature · 2023 · 619(7968):143-150doi:10.1038/s41586-023-06249-4 · PMID 37380764

    Mouse work with Novo Nordisk coauthors. On caloric restriction, GDF15 counteracted the compensatory fall in energy expenditure, giving greater weight loss and less fatty liver than caloric restriction alone, through a GFRAL and beta-adrenergic axis that raised fatty acid oxidation and calcium futile cycling in skeletal muscle. Animal only; it qualifies the 2017 conclusion that the hormone acts on intake alone.

45. 45
    Crawford J, Calle RA, Collins SM, et al. A Phase Ib First-In-Patient Study Assessing the Safety, Tolerability, Pharmacokinetics, and Pharmacodynamics of Ponsegromab in Participants with Cancer and Cachexia.Clin Cancer Res · 2024 · 30(3):489-497doi:10.1158/1078-0432.CCR-23-1631 · PMID 37982848

    Pfizer; 10 adults with advanced cancer, cachexia and raised GDF-15, open label, no control arm. Ponsegromab 200 mg subcutaneously every 3 weeks for 12 weeks; median unbound GDF-15 fell from 2.269 ng/mL to below the 0.0424 ng/mL limit of quantification from 3 hours after the first dose through week 15; least-squares mean weight rose 4.63 kg (SE 1.98) at week 12, about 6.6 percent; actigraphy and appetite scores improved; no treatment-related adverse events identified. Hypothesis-generating only, and the study the phase 2 trial was built on.

46. 46
    Fejzo M, Wang X, Tan Q, et al. Multi-ancestry genome-wide association study of severe pregnancy nausea and vomiting.Nat Genet · 2026 · 58(4):810-820doi:10.1038/s41588-026-02564-4 · PMID 41981316

    10,974 cases and 461,461 controls across European, Asian, African and Latino ancestries: ten loci, four previously known (GDF15, IGFBP7, PGR, GFRAL) and six new; GDF15 and TCF7L2 expressed mainly in extravillous trophoblasts, with GDF15 acting in opposite directions in the maternal and fetal genome.

47. 47
    Sharma R, Reinstadler B, Engelstad K, et al. Circulating markers of NADH-reductive stress correlate with mitochondrial disease severity.J Clin Invest · 2021 · 131(2):e136055doi:10.1172/JCI136055 · PMID 33463549

    Discovery and validation in 102 patients with the m.3243A>G variant and 32 controls: 20 analytes including GDF-15 separated patients from controls, correlated with functional status, and were nearly all attributable to a raised NADH to NAD+ ratio. Places GDF15 inside a defined biochemical derangement rather than treating it as a free-floating marker.

48. 48
    Kincaid JWR, Rimmington D, Tadross JA, et al. The gastrointestinal tract is a major source of the acute metformin-stimulated rise in GDF15.Sci Rep · 2024 · 14(1):1899doi:10.1038/s41598-024-51866-2 · PMID 38253650

    Tissue-specific deletion in mice: removing Gdf15 from the intestinal epithelium markedly reduced the circulating response to oral metformin while liver deletion did not; for the AMPK activator MK-8722 the liver mattered instead; intestinal GDF15 was not required for GDF15's restraint of high-fat-diet weight gain.

49. 49
    Zhang SY, Bruce K, Danaei Z, et al. Metformin triggers a kidney GDF15-dependent area postrema axis to regulate food intake and body weight.Cell Metab · 2023 · 35(5):875-886.e5doi:10.1016/j.cmet.2023.03.014 · PMID 37060902

    In high-fat-fed male rats, metformin raised plasma GDF15 by upregulating it in the kidney; kidney-specific knockdown of Gdf15, or knockdown of Gfral in the area postrema, abolished the effects on food intake and weight gain. Rats only, and males only.

50. 50
    Garrido AN, Zhang SY, Bruce K, Lai CSH, Yang Z, Wang MT, Lam TKT. Lipids Engage a Kidney-Brain GDF15 Axis to Suppress Food Intake.Diabetes · 2025 · 74(11):1918-1928doi:10.2337/db25-0174 · PMID 40877209

    Intralipid infused into the upper small intestine of chow-fed male rats lowered food intake and raised kidney, hepatic and plasma GDF15; kidney Gdf15 knockdown blunted both, and area postrema Gfral knockdown abolished the feeding effect. None of this occurred in high-fat-fed rats. Rats only.

51. 51
    Wang Q, Farhadipour M, Thijs T, et al. Bitter-tasting drugs tune GDF15 and GLP-1 expression via bitter taste or motilin receptors in the intestine of patients with obesity.Mol Metab · 2024 · 88:102002doi:10.1016/j.molmet.2024.102002 · PMID 39111389

    Placebo-controlled, double-blind, randomized two-visit crossover in healthy volunteers plus primary jejunal crypt cultures from patients with obesity. A single oral dose of hydroxychloroquine sulfate raised plasma GDF15, correlating with lower hunger scores and lower ghrelin; bitter compounds raised or lowered GDF15 expression through particular bitter taste receptor subtypes and the unfolded protein response, while azithromycin acted through the motilin receptor. Small, and the hunger correlation is not a weight outcome.

52. 52
    Conte M, Martucci M, Mosconi G, et al. GDF15 Plasma Level Is Inversely Associated With Level of Physical Activity and Correlates With Markers of Inflammation and Muscle Weakness.Front Immunol · 2020 · 11:915doi:10.3389/fimmu.2020.00915 · PMID 32477368

    228 people aged 18 to 83 in three groups: GDF15 rose with age, was lowest in amateur cyclists and highest in patients with lower-limb mobility impairment, rose sharply in cyclists immediately after a long-distance race, and correlated with quadriceps torque and leukocyte counts. Cross-sectional, so the direction of the activity relationship cannot be settled.

53. 53
    Johnson AA, Shokhirev MN, Wyss-Coray T, Lehallier B. Systematic review and analysis of human proteomics aging studies unveils a novel proteomic aging clock and identifies key processes that change with age.Ageing Res Rev · 2020 · 60:101070doi:10.1016/j.arr.2020.101070 · PMID 32311500

    Systematic review of 36 proteomic analyses: 1,128 proteins changed with age in two or more, and 32, including GDF15, in five or more; a proteomic clock predicted age in 3,301 people aged 18 to 76. Establishes GDF15 as one of the most reproducible age-associated proteins; association only.

54. 54
    Lopez-Cruz A, Figueredo Burgos NS, Hakimi AM, et al. Functional segregation of body-brain signals in the area postrema.bioRxiv · 2026 · preprint, 19 Jun 2026doi:10.64898/2026.06.15.732473 · PMID 42453677

    Preprint, not peer reviewed. Optical recordings in behaving mice: GFRAL neurons are activated by dietary fat independently of GDF15 and of canonical gut-brain pathways, and this activation is required for fat satiation; GIPR neurons, which directly inhibit GFRAL neurons, respond to sugar instead. Both are targets of tirzepatide. Mouse only, pending review.

55. 55
    Macia L, Tsai VW, Nguyen AD, et al. Macrophage inhibitory cytokine 1 (MIC-1/GDF15) decreases food intake, body weight and improves glucose tolerance in mice on normal and obesogenic diets.PLoS One · 2012 · 7(4):e34868doi:10.1371/journal.pone.0034868 · PMID 22514681

    Transgenic overexpression gave a lean phenotype on normal and high-fat diets, with reduced spontaneous but not fasting-induced food intake, unaltered energy expenditure, reduced physical activity and improved glucose tolerance. Mouse only.

56. 56
    Muscaritoli M, Molfino A, Orlando S, Tambaro F. Advancements of investigational agents for cancer cachexia: what clinical progress have we seen in the last 5 years?Expert Opin Investig Drugs · 2025 · 34(11):855-867doi:10.1080/13543784.2025.2588640 · PMID 41222020

    Independent review. Cachexia affects up to 80 percent of patients with advanced cancer; no FDA- or EMA-approved pharmacologic therapy exists; ponsegromab leads the pipeline and is entering phase 3, with AV-380 and NGM120 behind it; anamorelin and low-dose olanzapine are the non-GDF15 options.

57. 57
    Groarke JD, Crawford J, Collins SM, et al. Phase 2 study of the efficacy and safety of ponsegromab in patients with cancer cachexia: PROACC-1 study design.J Cachexia Sarcopenia Muscle · 2024 · 15(3):1054-1061doi:10.1002/jcsm.13435 · PMID 38500292

    The prespecified design: approximately 168 adults with lung, pancreatic or colorectal cancer, four arms, primary endpoint mean change in body weight at week 12 with a Bayesian analysis. Useful for confirming the endpoint was fixed in advance; contains no results.

58. 58
    Oswalt C, Crawford J. Novel Antibodies for Managing Cachexia.Annu Rev Med · 2026 · online 4 Aug 2026doi:10.1146/annurev-med-042325-042958 · PMID 42552253

    Review of the GDF-15/GFRAL axis as a primary mediator of cachexia and of the antibodies aimed at it, including other anti-GDF-15 and anti-GFRAL agents in development. Written by an author of the ponsegromab trial.

59. 59
    Breen DM, Joaquim S, Paulhus BL, et al. Targeting Cancer Cachexia: A Mechanistic Evaluation of Anti-GDF-15 Antibody-Based Combination Therapies.J Cachexia Sarcopenia Muscle · 2026 · 17(3):e70312doi:10.1002/jcsm.70312 · PMID 42126170

    Pfizer mouse work. Anti-myostatin antibody added to anti-GDF-15 helped only in a tumor model with low circulating GDF-15; anamorelin added weight but not significantly over anti-GDF-15 alone. Circulating myostatin correlated negatively with weight loss in patients in an advanced lung cancer study (NCT01360554). Mouse models, with one human correlation.

60. 60
    Dufour I, Bindels LB, Goffin E. Ponsegromab for the Treatment of Cancer Cachexia.N Engl J Med · 2025 · 392(10):1035doi:10.1056/NEJMc2500502 · PMID 40043243

    One of three letters contesting the ponsegromab trial, published March 2025. The PubMed record carries no abstract, so only its existence, not its substance, can be reported here.

61. 61
    Nielsen RL, Houlind MB, Andersen AL. Ponsegromab for the Treatment of Cancer Cachexia.N Engl J Med · 2025 · 392(10):1035-1036doi:10.1056/NEJMc2500502 · PMID 40043244

    A second letter on the ponsegromab trial, published March 2025. The PubMed record carries no abstract, so only its existence, not its substance, can be reported here.

62. 62
    Groarke JD, Sessa WC, Saxena AR. Ponsegromab for the Treatment of Cancer Cachexia. Reply.N Engl J Med · 2025 · 392(10):1036-1037doi:10.1056/NEJMc2500502 · PMID 40043245

    The trial authors' reply, from Pfizer, published March 2025. The PubMed record carries no abstract, so only its existence, not its substance, can be reported here.

63. 63
    Cohen AA, Picard M. Potential Risks of Blocking GDF15-Based Brain Energy Sensing.J Am Geriatr Soc · 2026 · 74(5):1516-1519doi:10.1111/jgs.70399 · PMID 41839791

    A short argument that GDF15 signals energetic stress to the brain so the body can conserve and reallocate energy, and that suppressing it may improve symptoms at the cost of long-term health and survival. A viewpoint, not data.

64. 64
    Melero I, de Miguel M, Cabanas EG, et al. Long-term follow-up of a phase 1/2 trial of anti-GDF-15 antibody visugromab plus anti-PD-1 antibody nivolumab in anti-PD-1/-L1 relapsed/refractory solid tumors.J Hematol Oncol · 2026 · 19(1)doi:10.1186/s13045-026-01818-2 · PMID 42426863

    Seventy-seven patients in three expansion cohorts: objective response in 4 of 22 non-squamous lung, 5 of 27 urothelial and 4 of 28 liver cancers, with median response durations of 32.2, 28.8 and 19.4 months and 7 of 13 responses ongoing. Single-arm and uncontrolled; the comparison with each patient's prior checkpoint therapy is within-patient, not randomized.

65. 65
    Melero I, de Miguel Luken M, de Velasco G, et al. Author Correction: Neutralizing GDF-15 can overcome anti-PD-1 and anti-PD-L1 resistance in solid tumours.Nature · 2025 · 639(8054):E18doi:10.1038/s41586-025-08827-0 · PMID 40000750

    The published author correction to the 2024 Nature paper. Noted here because corrections should be visible alongside the paper they amend; the PubMed record carries no abstract, so the substance of the correction is not quotable from it.

66. 66
    Filippini DM, Romaniello D, Carosi F, et al. The Multifaceted Role of Growth Differentiation Factor 15 (GDF15): A Narrative Review from Cancer Cachexia to Target Therapy.Biomedicines · 2025 · 13(8):1931doi:10.3390/biomedicines13081931 · PMID 40868185

    Narrative review of GDF15 across solid tumors: elevated levels associated with worse prognosis in colorectal, gastric, pancreatic, breast, lung, prostate and head and neck cancer, with laboratory evidence for roles in metastasis, immune evasion and chemoresistance. Narrative, not systematic, and the tumor-biology claims are largely associative or in vitro.

67. 67
    Kolnes KJ, Moller PM, Kruse R, et al. Metformin increases glycolysis and the stress-induced cytokine GDF15 but not FGF21 in humans.Front Endocrinol (Lausanne) · 2026 · 17:1797525doi:10.3389/fendo.2026.1797525 · PMID 41928890

    Randomized crossover in 34 healthy people, two 42-hour fasts with and without a week of metformin: serum GDF15 rose from 607 plus or minus 89 to 1,004 plus or minus 61 (p<0.001) while FGF21 did not change; the rise did not correlate with plasma metformin; Caco-2 intestinal cells secreted GDF15 and shifted to glycolysis. The abstract's units, ng/mL, must be a misprint for pg/mL. NCT01400191; Novo Nordisk coauthors.

68. 68
    Diaz M, Carreras-Badosa G, Villarroya J, et al. Circulating GDF15 concentrations in girls with low birth weight: effects of prolonged metformin treatment.Pediatr Res · 2022 · 93(4):964-968doi:10.1038/s41390-022-02175-9 · PMID 35817957

    Thirty girls born small with rapid catch-up and early pubic hair, randomized to four years of metformin or no treatment: GDF15 rose significantly only in years three and four, and correlated inversely with insulin, HOMA-IR, androgens, body fat and visceral fat. Very small, and the GDF15 mediation is inferred, not shown.

69. 69
    de Zegher F, Diaz M, Villarroya J, et al. The relative deficit of GDF15 in adolescent girls with PCOS can be changed into an abundance that reduces liver fat.Sci Rep · 2021 · 11(1):7018doi:10.1038/s41598-021-86317-9 · PMID 33782413

    Post-hoc endpoint added to two randomized studies in 58 non-obese adolescent girls with polycystic ovary syndrome: an oral contraceptive nearly doubled GDF15 (296 to 507 pg/mL) while also raising CRP, whereas a low-dose spironolactone-pioglitazone-metformin combination raised it 3.4-fold (308 to 1045 pg/mL) and lowered CRP, insulin and liver fat without changing body weight. Post-hoc, and small.

70. 70
    Hoekx CA, Straat ME, Bizino MB, et al. Growth differentiation factor 15 is not modified after weight loss induced by liraglutide in South Asians and Europids with type 2 diabetes mellitus.Exp Physiol · 2024 · 109(8):1292-1304doi:10.1113/EP091815 · PMID 38965822

    Twenty-six weeks of liraglutide (44 patients) versus placebo (50) in type 2 diabetes: weight fell, plasma GDF15 did not change in either ethnic group, and changes in GDF15 were unrelated to changes in weight or HbA1c. A clean negative for the idea that GLP-1 drugs work through this hormone.

71. 71
    Fiorenza M, Checa A, Sandsdal RM, et al. Weight-loss maintenance is accompanied by interconnected alterations in circulating FGF21-adiponectin-leptin and bioactive sphingolipids.Cell Rep Med · 2024 · 5(7):101629doi:10.1016/j.xcrm.2024.101629 · PMID 38959886

    Secondary analysis of a randomized trial: adults with obesity through diet-induced weight loss and then a maintenance phase with exercise, a GLP-1 analogue, both or placebo. GDF15 and C16:0 to C18:0 ceramides rose transiently during the initial low-calorie diet phase. It means a drug trial that measures GDF15 after weight has already fallen may be sampling the wrong moment.

72. 72
    Petry CJ, Ong KK, Burling KA, et al. Associations of vomiting and antiemetic use in pregnancy with levels of circulating GDF15 early in the second trimester: A nested case-control study.Wellcome Open Res · 2018 · 3:123doi:10.12688/wellcomeopenres.14818.1 · PMID 30345390

    791 women, blood at a median of 15.1 weeks: GDF15 was 11,581 pg/mL in the 175 who vomited in the second trimester against 10,593 in the 193 with no nausea or vomiting (p=0.02), and 13,157 in the 11 taking antiemetics; levels correlated positively with hCG and inversely with maternal BMI. Observational, one time point, modest separation.

73. 73
    Svenningsen JS, Ostenfeld A, Petersen JF, et al. Early pregnancy GDF15 trajectories and their association with normal gestation, pregnancy loss, and hyperemesis gravidarum.EBioMedicine · 2026 · 131:106428doi:10.1016/j.ebiom.2026.106428 · PMID 42561814

    Longitudinal serum sampling with PUQE-24 scores from two early-pregnancy cohorts: GDF15 rose until week 7 and plateaued by week 9; female fetuses ran about 19 percent higher and male fetuses rose a day later; symptom scores peaked around week 8; in hyperemesis the hormone had plateaued by diagnosis, with higher early levels but overlapping intervals by week 8; early pregnancy loss showed lower, slower rises. Defines sampling windows rather than a diagnostic cut-off.

74. 74
    Sharma N, MacGibbon KW, Brecht-Doscher A, Cortessis VK, Fejzo MS. Prepregnancy metformin use associated with lower risk of severe nausea and vomiting of pregnancy and hyperemesis gravidarum.Am J Obstet Gynecol · 2025 · 233(6):649.e1-649.e14doi:10.1016/j.ajog.2025.06.055 · PMID 40588059

    Questionnaire study of 5,414 women recruited through a hyperemesis patient organization: metformin in the month before a first pregnancy carried an adjusted relative risk of 0.29 (0.12-0.71) for hyperemesis, and 0.18 (0.06-0.59) before a second pregnancy; tobacco 0.51; SSRIs 2.41; recurrence 86 percent. Self-reported, retrospective, and recruited from a population enriched for the condition.

75. 75
    Sillis L, Heinonen EW, Ceulemans M, Johnson D, Luo Y, Chambers CD. Metformin for the Prevention of Hyperemesis Gravidarum: An Observational Cohort Study.BJOG · 2025 · 132(12):1772-1778doi:10.1111/1471-0528.18238 · PMID 40443168

    MotherToBaby cohort: hyperemesis in 1 of 80 women exposed to metformin before conception against 97 of 4,411 unexposed, adjusted relative risk 0.50 with a confidence interval from 0.07 to 3.39; rates and duration of ordinary nausea and vomiting were the same. Explicitly underpowered.

76. 76
    Sharma A, Hijazi Z, Andersson U, et al. Use of Biomarkers to Predict Specific Causes of Death in Patients With Atrial Fibrillation.Circulation · 2018 · 138(16):1666-1676doi:10.1161/CIRCULATIONAHA.118.034125 · PMID 29871978

    ARISTOTLE biomarker substudy, 14,798 patients, 1,272 deaths. A doubling of GDF-15 was most strongly associated with death from bleeding (hazard 1.72, p=0.028), while troponin T was strongest for sudden death and natriuretic peptide for heart failure death. Observational within a randomized trial.

77. 77
    Hijazi Z, Lindback J, Oldgren J, et al. Evaluation of the updated ABC-AF-bleeding score 2.0 in patients with atrial fibrillation treated with a direct oral anticoagulant or warfarin.J Thromb Haemost · 2026 · 24(2):399-407doi:10.1016/j.jtha.2025.09.032 · PMID 41644239

    25,962 patients pooled from three pivotal anticoagulation trials, 1,321 major bleeds. Version 2.0 adds the type of oral anticoagulant (direct oral anticoagulant or warfarin) to the model, alongside age, bleeding history and the biomarkers GDF-15, hemoglobin and troponin T; C-index 0.69 for major bleeding, 0.72 gastrointestinal, 0.66 intracranial, better than HAS-BLED, ORBIT and the DOAC score.

78. 78
    Schlegl J, Bannehr M, Lichtenauer M, et al. Biomarkers of systemic disease burden and outcomes after transcatheter tricuspid edge-to-edge repair.BMC Cardiovasc Disord · 2026 · 26(1)doi:10.1186/s12872-026-06392-6 · PMID 42547888

    Prospective single-center cohort of 84 consecutive patients; 10 deaths and 29 cardiovascular rehospitalizations at 12 months. GDF-15 discriminated all-cause death with an area under the curve of 0.823 (0.714 to 0.932) against 0.535 (0.362 to 0.709) for NT-proBNP, which did not discriminate; soluble urokinase plasminogen activator receptor was intermediate at 0.742. The authors call all analyses exploratory, and 10 events is very few.

79. 79
    Shayota BJ. Biomarkers of mitochondrial disorders.Neurotherapeutics · 2024 · 21(1):e00325doi:10.1016/j.neurot.2024.e00325 · PMID 38295557

    Review of thirteen candidate biomarkers. Concludes that in large comparison studies GDF-15, followed by FGF-21, has the greatest value, while stating that neither is adequate alone and that findings are mixed across specific diseases.

80. 80
    Wang X, Chrysovergis K, Kosak J, et al. hNAG-1 increases lifespan by regulating energy metabolism and insulin/IGF-1/mTOR signaling.Aging (Albany NY) · 2014 · 6(8):690-704doi:10.18632/aging.100687 · PMID 25239873

    Female mice carrying a human GDF15 transgene had longer mean and median lifespans in two independent lines, with a larger effect on a high-fat diet; they were leaner, with lower IGF-1, insulin and glucose, better insulin sensitivity and reduced IGF-1/mTOR signaling. A genuine mouse lifespan study in females; it says nothing directly about humans.

81. 81
    Justesen JM, Sorensen KV, Seshadri J, et al. A loss-of-function variant in GFRAL associates with increased alcohol consumption in humans.bioRxiv · 2026 · preprint, 9 Mar 2026doi:10.64898/2026.03.06.709997 · PMID 41959486

    Preprint, not peer reviewed. Reports circulating GDF15 about fivefold higher in alcohol dependence, a rare loss-of-function GFRAL variant associated with about 2.6 extra UK alcohol units (roughly 21 g ethanol) per week, and reduced alcohol drinking in mice given recombinant GDF15. A hypothesis with human genetic and mouse support, pending review.

82. 82
    Najjar NR, De Jonghe BC, Hayes MR, Doyle RP. From stress cytokine to drug target: The translational landscape of growth differentiation factor 15.J Pharmacol Exp Ther · 2026 · 393(11):105036doi:10.1016/j.jpet.2026.105036 · PMID 42880196

    Current review of GDF15-GFRAL-RET biology from a medicinal chemistry standpoint, covering agonists, antagonists and circulating-ligand traps, patents and ongoing trials on both sides of the axis.

83. 83
    Li BUK. Landscape of Cyclic Vomiting Syndrome: From Bedside to Bench, Past to Present.Neurogastroenterol Motil · 2025 · 37(3):e14990doi:10.1111/nmo.14990 · PMID 39789960

    Review of cyclic vomiting syndrome and cannabinoid hyperemesis syndrome. Lists pathways the author counts as confirmed (the substance P receptor NK1, the cannabinoid receptor CB1, the hypothalamic-pituitary-adrenal axis, the PPM1D gene, estrogen) and possible ones, GDF-15 among them with TRPV-1, calcitonin gene-related peptide and mitochondrial dysfunction. A hypothesis list, not human GDF15 data.