{
  "id": "gadolinium-based-contrast-agent",
  "title": "Gadolinium-Based Contrast Agent",
  "lang": "en",
  "source_lang": "en",
  "summary": "Gadolinium-based contrast agents are chelated metal compounds that alter magnetic resonance signals to improve tissue visualization and diagnostic assessment.",
  "category": "health",
  "subcategory": "drugs",
  "type": "concept",
  "aliases": [
    "Gadolinium contrast agent",
    "Gadolinium-based contrast medium",
    "GBCA",
    "Gadolinium chelate"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/gadolinium-based-contrast-agent/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/gadolinium-based-contrast-agent/",
      "json": "https://aiwiki.page/zh/gadolinium-based-contrast-agent.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A gadolinium-based contrast agent (GBCA) is a pharmaceutical preparation used to improve the visibility of tissues, blood vessels, and abnormalities in magnetic resonance imaging (MRI). Its active component contains gadolinium, a strongly paramagnetic metal, bound within an organic ligand that greatly reduces the toxicity associated with the unbound ion. Unlike the iodinated agents commonly used in computed tomography, GBCAs act principally by changing the magnetic relaxation of nearby water protons rather than by absorbing X-rays.\n\nThese agents are important in the assessment of tumors, vascular disorders, inflammatory disease, and selected abnormalities of the liver, heart, and other organs. Their effects depend not only on the administered compound but also on blood flow, vascular permeability, extracellular space, organ function, and the MRI acquisition technique. Enhancement is therefore a physiological signal, not a disease-specific label.\n\nSince the first clinical approvals in 1988, GBCA development has involved a balance between diagnostic performance and chemical stability. Major safety questions concern acute hypersensitivity reactions, nephrogenic systemic fibrosis in susceptible patients, and the retention of small amounts of gadolinium after administration. Differences in molecular structure, elimination, and clinical evidence make individual agents distinct rather than interchangeable members of a uniform class."
    },
    {
      "heading": "Chemical composition and major classes",
      "text": "Gadolinium is a lanthanide element with atomic number 64. In contrast agents it is normally present as the trivalent ion, \\(\\mathrm{Gd}^{3+}\\), whose seven unpaired electrons produce a large magnetic moment. The unbound ion is unsuitable for routine intravenous administration because it can interfere with biological processes involving calcium and other metal ions. Clinical agents therefore use chelation: a ligand coordinates the metal through several donor atoms, producing a complex that is substantially more stable and biologically manageable than a simple gadolinium salt.\n\nThe principal structural distinction is between linear and macrocyclic complexes. A linear ligand wraps around the metal in an open-chain arrangement. Examples include gadopentetate dimeglumine, gadodiamide, gadoversetamide, gadobenate dimeglumine, and gadoxetate disodium. A macrocyclic ligand forms a ring-like cage around the metal. Gadoterate meglumine, gadobutrol, gadoteridol, and gadopiclenol belong to this group. The cage generally makes release of gadolinium kinetically more difficult, although the detailed behavior of each compound depends on its complete coordination chemistry.\n\nA second distinction is between ionic and nonionic agents. This describes the charge of the gadolinium-containing complex, not whether the injectable solution contains any ions. Ionic complexes are formulated with counterions, such as meglumine or sodium, whereas nonionic complexes do not require a counterion to balance the complex itself. Ionicity affects formulation properties, including osmolality, but does not by itself determine clinical safety. For example, both gadobutrol and gadodiamide are nonionic, yet their macrocyclic and linear structures give them markedly different stability characteristics.\n\nTwo measures of stability require separation. Thermodynamic stability describes the equilibrium tendency of a complex to remain intact under specified conditions. Kinetic inertness describes how slowly it dissociates. A complex can have a favorable equilibrium constant but still release metal more rapidly than another agent under biological conditions. Competition from endogenous metals, ligand protonation, and the prolonged residence time associated with poor elimination can influence these processes. Exchange of the coordinated metal with another metal is called transmetallation.\n\nThe administered formulation also contains components that maintain pH and pharmaceutical stability; some preparations include excess ligand or related stabilizing compounds. Consequently, a product is not simply elemental gadolinium dissolved in water. Its safety and imaging behavior arise from the metal–ligand complex, its formulation, and its interaction with the body. Structural labels such as “linear” and “macrocyclic” are useful, but they do not replace agent-specific evidence or regulatory labeling."
    },
    {
      "heading": "Magnetic relaxation and the production of image contrast",
      "text": "The principal action of a GBCA is to accelerate the relaxation of nearby hydrogen nuclei after radiofrequency excitation. In MRI, longitudinal relaxation, described by \\(T_1\\), concerns recovery of magnetization along the main magnetic field. Transverse relaxation, described by \\(T_2\\), concerns loss of phase coherence in the transverse plane. Gadolinium affects both processes. At concentrations and imaging settings commonly used for clinical enhancement, shortening of \\(T_1\\) usually predominates in the visible result, making exposed tissues appear brighter on appropriately weighted images.\n\nThe effect on longitudinal relaxation can often be approximated by\n\n\\[R_1 = R_{1,0} + r_1 C,\\]\n\nwhere \\(R_1=1/T_1\\), \\(R_{1,0}\\) is the relaxation rate before contrast administration, \\(C\\) is the gadolinium concentration, and \\(r_1\\) is longitudinal relaxivity. Relaxivity is commonly expressed in \\(\\mathrm{L\\,mmol^{-1}\\,s^{-1}}\\). A corresponding relationship can be written for transverse relaxation. These equations describe relaxation rates under specified conditions; they do not mean that image brightness is directly proportional to contrast concentration.\n\nRelaxation enhancement depends on several molecular processes. Water may coordinate directly to the gadolinium ion and exchange with surrounding water, while other water molecules interact at greater distances. The number of coordinated water molecules, their exchange rate, molecular rotation, and electronic relaxation all contribute. Larger molecules or agents that bind to plasma proteins may rotate more slowly and show increased relaxivity at some field strengths. The relationship is not universal: relaxivity changes with magnetic field, temperature, and the biological environment.\n\nAt high local concentrations, transverse relaxation and magnetic susceptibility effects can dominate, producing signal loss rather than brightening. This occurs, for example, during the first passage of a contrast bolus through cerebral vessels in susceptibility-based perfusion imaging. Concentrated contrast in the urinary tract can also produce complex signal behavior. The effective transverse relaxation time, \\(T_2^*\\), includes contributions from local magnetic-field inhomogeneity and is particularly important in these applications.\n\nAcquisition parameters determine how these physical effects become an image. Repetition time, echo time, flip angle, spatial resolution, and fat suppression all influence the appearance of enhancement. Comparisons usually depend on appropriately matched precontrast and postcontrast sequences. GBCAs are therefore signal modifiers rather than dyes whose presence can be interpreted independently of the instrument. They also differ fundamentally from radioactive tracers: routine gadolinium-enhanced MRI uses neither radioactive decay nor ionizing radiation to generate its signal."
    },
    {
      "heading": "Distribution, elimination, and tissue enhancement",
      "text": "Most conventional GBCAs behave as small extracellular molecules. Following intravenous injection, they initially circulate in plasma and then distribute into accessible extracellular fluid. They ordinarily do not enter intact cells in large amounts. Their pharmacokinetics therefore reflect vascular delivery, movement across capillary walls, extracellular distribution volume, and elimination. In adults with normal renal function, many extracellular agents have a plasma elimination half-life of approximately 1.5–2 hours, although exact values vary among products and populations.\n\nThe blood–brain barrier strongly limits entry of conventional extracellular agents into normal brain tissue. Enhancement within a brain lesion commonly indicates altered barrier permeability, abnormal vasculature, or both. Tumors, active inflammatory lesions, and some infections can create such changes. Conversely, the absence of enhancement does not establish normality: important abnormalities may preserve the barrier or have physiological characteristics that produce little visible contrast accumulation. Structures outside the conventional barrier, including parts of the pituitary region and choroid plexus, may normally enhance.\n\nRenal elimination occurs principally through glomerular filtration for most extracellular agents. Clearance consequently falls as glomerular filtration rate declines, and circulating exposure can be prolonged substantially in advanced renal failure. The administered compound is generally excreted without requiring extensive metabolism. This explains why kidney function is central to GBCA safety assessment even when the organ being imaged is elsewhere in the body.\n\nSome agents have additional hepatobiliary behavior. Gadoxetate disodium is taken up by functioning hepatocytes through organic anion transport pathways and is subsequently secreted into bile; in people with normal organ function, renal and hepatobiliary elimination each account for approximately half the dose. Gadobenate dimeglumine has a much smaller hepatobiliary fraction in humans, generally a few percent. Such differences create distinct imaging windows and cannot be inferred merely from whether an agent is linear or macrocyclic.\n\nEnhancement also changes with time. Early images may emphasize arterial delivery, subsequent images may depict venous circulation and extracellular distribution, and later images may demonstrate persistent interstitial accumulation or hepatocyte uptake. Delayed enhancement in myocardial scar, for example, reflects altered extracellular distribution and washout rather than specific binding to scar tissue. The physiological interpretation of enhancement thus depends on the agent, tissue, disease process, and interval between injection and image acquisition."
    },
    {
      "heading": "Diagnostic applications and interpretation",
      "text": "In radiology, GBCAs are used when their effects supply information that noncontrast sequences do not adequately provide. Their major contributions include identifying altered vascular permeability, distinguishing enhancing tissue from fluid or necrosis, characterizing vascular supply, and mapping the extent of disease. They are not required for every MRI examination. Many questions involving joints, spinal degeneration, acute ischemia, and selected vascular conditions can be addressed with noncontrast techniques, depending on the specific clinical setting.\n\nIn the central nervous system, contrast enhancement helps characterize primary tumors, metastases, meningeal disease, and some infectious and inflammatory abnormalities. In multiple sclerosis, an enhancing lesion indicates blood–brain barrier disruption associated with recent inflammatory activity, whereas a nonenhancing lesion may represent an older or otherwise inactive focus. Enhancement patterns contribute to interpretation but are not individually diagnostic: ring enhancement, for example, can occur in tumors, abscesses, and demyelinating lesions.\n\nIn oncological imaging outside the nervous system, dynamic enhancement provides information about vascular delivery and tissue composition. Breast MRI uses enhancement kinetics together with morphology to assess suspicious lesions and disease extent. Multiparametric prostate MRI may incorporate dynamic contrast enhancement alongside anatomical imaging and diffusion MRI. In the liver, arterial, portal venous, delayed, and hepatobiliary phases can help distinguish lesion types. A dark lesion in a hepatobiliary phase may indicate absent or reduced functioning hepatocytes, but that finding alone does not establish malignancy.\n\nCardiovascular applications use several different mechanisms. Contrast-enhanced magnetic resonance angiography depicts arterial or venous lumina during passage of a timed bolus. Myocardial perfusion imaging evaluates first-pass delivery, while late gadolinium enhancement identifies characteristic patterns of tissue injury or expanded extracellular space. After myocardial infarction, enhancement commonly begins in the subendocardium and may extend through the wall in a coronary distribution. Other cardiac diseases can produce different patterns, requiring interpretation alongside ventricular function and other tissue measurements.\n\nQuantitative approaches attempt to move beyond visual brightness. Dynamic contrast-enhanced MRI can estimate parameters related to vascular permeability and contrast transfer, such as \\(K^{\\mathrm{trans}}\\), but these estimates depend on the model, arterial input function, acquisition, and physiological assumptions. Cardiac extracellular-volume measurement combines changes in tissue and blood relaxation with hematocrit. Such measurements can function as imaging biomarkers, yet they are not automatically equivalent to direct measurements of a single biological property. Perfusion, permeability, extracellular volume, and technical artifacts can contribute simultaneously to the observed signal."
    },
    {
      "heading": "Administration, dose, and technical performance",
      "text": "Clinical GBCAs are usually administered intravenously, either manually or with a power injector. A saline flush can move the injected bolus through the connecting tubing and peripheral veins, helping produce a more reproducible arrival profile. Injection rate and scan timing vary with the examination: angiography and first-pass perfusion depend heavily on bolus shape, whereas many delayed tissue-enhancement studies are less sensitive to the precise initial concentration curve. These are product- and protocol-specific technical considerations rather than uniform properties of the drug class.\n\nDose is expressed as millimoles of gadolinium per kilogram of body mass, not simply as milliliters of solution. Many extracellular agents have a conventional labeled dose of \\(0.1\\ \\mathrm{mmol/kg}\\), but important exceptions exist. Gadoxetate disodium is commonly administered at \\(0.025\\ \\mathrm{mmol/kg}\\), while gadopiclenol has a labeled dose of \\(0.05\\ \\mathrm{mmol/kg}\\) for its approved indications. These differences reflect distinct molecular properties and evidence from product development; they do not establish a universal dose-equivalence rule.\n\nSolution concentration must be distinguished from administered dose. Gadobutrol is formulated at \\(1.0\\ \\mathrm{mmol/mL}\\), whereas many other products are formulated at \\(0.5\\ \\mathrm{mmol/mL}\\); gadoxetate is commonly supplied at \\(0.25\\ \\mathrm{mmol/mL}\\). The relationship is\n\n\\[V=\\frac{D\\,m}{C_s},\\]\n\nwhere \\(V\\) is injection volume, \\(D\\) is the prescribed amount per kilogram, \\(m\\) is body mass, and \\(C_s\\) is solution concentration. A smaller injection volume therefore does not necessarily mean a smaller gadolinium dose.\n\nTechnical failures can change diagnostic appearance. Motion between precontrast and postcontrast acquisitions can produce false findings on subtraction images. Mistimed arterial acquisition may obscure vessels or lesions whose enhancement is brief. Some patients receiving gadoxetate experience transient respiratory motion that particularly affects arterial-phase liver images. Contrast extravasation into soft tissue usually causes localized symptoms rather than systemic toxicity, but its severity depends on the circumstances and volume involved. These events are distinct from immunological reactions.\n\nIntrathecal administration is not a routine approved route for these agents and has been associated with serious neurotoxicity, particularly after dosing errors. Likewise, direct joint administration for magnetic resonance arthrography involves a different, highly diluted local exposure, with regulatory status varying by product and jurisdiction. Safety evidence from standard intravenous use cannot simply be transferred to other routes. Product identification, dose units, concentration, and route are consequently separate elements of patient safety in contrast-enhanced imaging."
    },
    {
      "heading": "Acute adverse reactions and special populations",
      "text": "Acute adverse events include both physiological effects and allergic-like reactions. Transient warmth, altered taste, nausea, and injection-site discomfort may occur without an immune mechanism. Allergic-like manifestations include urticaria, facial swelling, bronchospasm, and, rarely, severe cardiovascular or respiratory compromise. Across commonly cited clinical series, allergic-like reactions occur in approximately \\(0.01\\%\\)–\\(0.22\\%\\) of administrations, with severe reactions around \\(0.008\\%\\); estimates differ by agent, population, and reporting method.\n\nThe term “allergic-like” reflects biological heterogeneity. Some reactions involve specific immunological mechanisms, while others may arise through non-IgE pathways that activate mast cells or related mediators. Anaphylaxis is a clinical syndrome and does not require proof of a particular antibody pathway at the time of the event. A previous reaction to a GBCA is an important risk factor for another reaction. A reaction to an iodinated contrast agent does not establish chemical cross-reactivity with a GBCA, although some patients have a broader predisposition to contrast reactions.\n\nImmediate hypersensitivity and delayed reactions require distinction. Immediate events usually arise during administration or within a short interval afterward. Delayed skin eruptions have also been described, but they are less well characterized than immediate reactions. Attribution can be difficult because patients undergoing imaging may receive other medications or have concurrent illnesses. Pharmacovigilance therefore depends on accurate documentation of the specific product, timing, manifestations, and other relevant exposures rather than the nonspecific label “contrast allergy.”\n\nDuring pregnancy, gadolinium complexes can cross the placenta, enter the fetal circulation, and reach amniotic fluid through fetal renal excretion. Human evidence is limited and largely observational. Regulatory labels and professional guidance therefore place restrictions on use during pregnancy, generally reserving it for situations in which the expected diagnostic benefit is important and cannot be obtained adequately without contrast. These restrictions reflect uncertainty about fetal exposure rather than proof that every exposure causes harm.\n\nOnly a small fraction of a maternal intravenous dose appears in breast milk, and gastrointestinal absorption of the ingested complex by the infant is also low. Professional guidance generally regards breastfeeding interruption as unnecessary after standard intravenous administration. Pediatric use has additional product-specific considerations, including approved age ranges, weight-based dosing, and immature renal function in neonates. Evidence and labeling are not identical across agents, so statements about adult use do not automatically establish approval for infants or children."
    },
    {
      "heading": "Kidney dysfunction and nephrogenic systemic fibrosis",
      "text": "The most consequential established delayed toxicity associated with certain GBCAs is nephrogenic systemic fibrosis (NSF). First described as a distinct disorder in 2000, it was initially called nephrogenic fibrosing dermopathy. It causes thickening and hardening of the skin and subcutaneous tissues and may produce painful contractures and severe loss of mobility. Internal organs can also be involved. Its association with gadolinium exposure in patients with severe renal dysfunction was recognized in 2006.\n\nMost historical cases occurred in people with advanced chronic kidney disease, acute kidney injury, or dependence on dialysis. A glomerular filtration rate below approximately \\(30\\ \\mathrm{mL/min/1.73\\,m^2}\\) identifies an important risk category, although risk is not determined by one numerical threshold alone. Acute illness, repeated exposures, dose, and the particular agent also matter. Many early cases followed relatively high or repeated doses of agents subsequently recognized as having greater NSF risk.\n\nThe proposed mechanism involves prolonged exposure to less stable gadolinium complexes, release or transformation of gadolinium-containing species, and activation of inflammatory and profibrotic pathways. The exact sequence is not completely resolved. Tissue gadolinium can be detected in affected patients, but deposition alone does not explain every aspect of the disease. The relationship between exposure, renal dysfunction, and agent type is nevertheless sufficiently established to support major changes in clinical practice.\n\nRisk classifications used by professional organizations distinguish agents according to observed NSF associations rather than structure alone. In the American College of Radiology framework, group I includes gadodiamide, gadopentetate dimeglumine, and gadoversetamide, which account for most historically implicated exposures. Group II includes agents with few, if any, unconfounded cases, including several macrocyclic agents and the linear agent gadobenate dimeglumine. Group membership can change as evidence accumulates, so it is not synonymous with a permanent chemical classification.\n\nA 2020 meta-analysis covering 4,931 patients with stage 4 or 5 chronic kidney disease found no NSF cases after group II exposure; the upper bound of the pooled 95% confidence interval was \\(0.07\\%\\). This indicates a very low estimated risk in the studied settings, not mathematical proof of zero risk. Following restrictions on higher-risk agents and changes in dosing and screening, new NSF reports declined markedly.\n\nHemodialysis removes circulating GBCA, but removal has not been shown to guarantee prevention of NSF. Contemporary consensus statements do not support initiating or intensifying dialysis solely because a standard dose of a low-NSF-risk agent has been administered. NSF is also distinct from acute kidney injury attributed to a contrast exposure: at approved intravenous doses, GBCAs do not show the same nephrotoxicity profile associated with some other contrast settings."
    },
    {
      "heading": "Gadolinium retention and its clinical significance",
      "text": "Small amounts of gadolinium can remain in the body after most of an administered dose has been eliminated. Retention has been demonstrated in bone, skin, brain, and other tissues, including in people without severe renal impairment. Bone may serve as a comparatively long-lived reservoir. The amount retained varies with cumulative exposure, molecular structure, time since administration, organ function, and the analytical method used. “Retention” describes persistence; it does not by itself establish tissue injury.\n\nThe issue became especially prominent after a 2014 report linked repeated GBCA exposure to increased signal on unenhanced \\(T_1\\)-weighted images in the dentate nucleus and globus pallidus. Subsequent studies, including tissue analyses, confirmed that gadolinium could be present in these regions. Signal changes have been reported more consistently after repeated exposure to certain linear agents than after exposure to macrocyclic agents. However, a visible MRI signal change is neither necessary nor sufficient to measure all retained gadolinium.\n\nAnalytical methods answer different questions. Inductively coupled plasma mass spectrometry can measure total elemental gadolinium at very low concentrations, but it does not on its own establish whether the metal remains in its original chelate. Chemical-speciation studies investigate intact complexes, soluble macromolecule-associated forms, and less soluble compounds. Animal studies show that the proportions and clearance of these species differ among agents. This distinction matters because equal elemental amounts need not have identical biological effects.\n\nMacrocyclic agents generally show lower tissue retention than linear agents under comparable conditions, but “macrocyclic” does not mean complete absence of retention. Conversely, detecting gadolinium after a linear agent does not establish that clinical damage has occurred. Evaluations of possible neurological effects must account for the underlying disease, other treatments, age, exposure history, and the limitations of retrospective data. Repeated MRI is itself more common in people with neurological or oncological illness, creating substantial potential for confounding.\n\nRegulators have not established a causal relationship between brain gadolinium retention and a specific neurological disorder in patients with normal renal function. Reports of persistent multisystem symptoms after exposure have prompted research and regulatory attention, but they do not yet define a uniformly accepted causal disease entity with a validated diagnostic test. Proposed terminology such as “gadolinium deposition disease” should therefore be distinguished from NSF, whose clinical association is established.\n\nMeasurements of gadolinium in urine or blood can document exposure or ongoing elimination in some circumstances, but they do not independently diagnose toxicity. No broadly accepted threshold separates harmless retention from clinically significant injury. The unresolved questions concern chemical form, tissue localization, long-term biological response, and whether particular susceptible populations experience effects not captured by existing studies."
    },
    {
      "heading": "Development, regulation, and continuing research",
      "text": "The clinical development of GBCAs followed the emergence of MRI as a diagnostic technology in the late 1970s and early 1980s. Investigators needed a compound that combined strong relaxation effects with acceptable biological behavior. Gadopentetate dimeglumine became an early widely used solution and received United States approval in 1988. Subsequent development expanded the range of extracellular agents, introduced macrocyclic alternatives, and produced compounds with partial hepatobiliary uptake or other specialized properties.\n\nThe NSF experience changed both regulation and clinical research. From 2006 onward, safety communications increasingly emphasized renal impairment, agent selection, and exposure history. In 2007, the United States Food and Drug Administration required boxed warnings concerning NSF for GBCAs, with later revisions and additional restrictions for higher-risk products. The episode demonstrated that safety characteristics could not be adequately represented by treating all gadolinium compounds as equivalent.\n\nRetention prompted a second major regulatory response. In 2017, European regulators recommended suspension of several intravenous linear agents while retaining selected products for particular indications, including certain liver examinations. The United States did not adopt the same broad suspensions, but the FDA required class warnings and patient information addressing gadolinium retention. These differing responses reflected different regulatory judgments about precaution, diagnostic alternatives, and the available evidence, rather than a disagreement that retention occurs.\n\nA major development objective is obtaining adequate diagnostic information with less administered gadolinium. Gadopiclenol, first approved in the United States in 2022, is a macrocyclic agent designed for high relaxivity and is used at a lower molar dose than many conventional extracellular agents. Higher relaxivity can support dose reduction in validated applications, but diagnostic equivalence must be established through appropriate clinical trials. Dose reduction cannot be assumed from a laboratory relaxivity measurement alone.\n\nOther research directions include targeted complexes, protein-binding agents, macromolecular carriers, and probes intended to respond to enzymes or particular tissue environments. Such designs face additional challenges: stronger tissue binding or longer circulation may improve imaging while also prolonging exposure. Alternative paramagnetic elements, including manganese and iron, are also investigated, but each has distinct chemistry, imaging behavior, and toxicological constraints.\n\nNoncontrast imaging provides a parallel route to reducing exposure. Arterial spin labeling uses magnetically labeled blood water to estimate perfusion; diffusion imaging assesses water motion; and several angiographic techniques exploit flow without injected contrast. Computational reconstruction and machine learning can improve image acquisition or generate estimated contrast-like appearances, but a synthesized image does not necessarily preserve every biological feature revealed by an actual GBCA. Validation must address missed lesions, altered enhancement patterns, generalizability, and performance in the specific clinical population."
    }
  ],
  "links": [
    "magnetic-resonance-imaging",
    "x-ray-computed-tomography",
    "chelation",
    "magnetic-field",
    "protein",
    "pharmacokinetics",
    "blood-brain-barrier",
    "radiology",
    "central-nervous-system",
    "metastasis",
    "multiple-sclerosis",
    "diffusion-mri",
    "hematocrit",
    "biomarker",
    "patient-safety",
    "nausea",
    "urticaria",
    "anaphylaxis",
    "pharmacovigilance",
    "pregnancy",
    "nephrogenic-systemic-fibrosis",
    "chronic-kidney-disease",
    "clinical-trial",
    "machine-learning"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}