AUC

AUC
Name	AUC
Field	Statistics; Pharmacology

AUC

Area under the curve (AUC) denotes the integral of a function between specified limits and serves as a scalar summary of curve magnitude. In statistical and pharmacokinetic contexts, it quantifies aggregate signal, exposure, or performance across a continuous domain. AUC is widely applied in evaluation tasks across disciplines associated with George Box, Jerzy Neyman, Edwin G. Boring, Karl Pearson and institutions such as Harvard University, University of Cambridge, Stanford University, Massachusetts Institute of Technology.

Definition and terminology

AUC commonly refers to the definite integral ∫ f(x) dx over an interval, described by mathematical analysis developed by Augustin-Louis Cauchy and formalized in modern measure theory by Henri Lebesgue and Émile Borel. In binary classification, AUC denotes the area under the receiver operating characteristic curve, which plots true positive rate versus false positive rate; this formulation traces conceptual roots to signal detection work by John A. Swets and operational research at Bell Labs. Pharmacokinetic AUC represents drug plasma concentration integrated over time and is central to regulatory guidance by agencies such as the Food and Drug Administration and the European Medicines Agency.

Mathematical properties and computation

As an integral, AUC is linear: AUC(a,b)[f+g] = AUC(a,b)[f] + AUC(a,b)[g]; scalar multiplication holds via properties proved in texts by Walter Rudin and Michael Spivak. Numerical estimation employs quadrature rules including trapezoidal rule linked to methods in Carl Friedrich Gauss's quadrature and Simpson's rule associated with Thomas Simpson. For ROC AUC, equivalence with the probability that a randomly chosen positive instance ranks higher than a randomly chosen negative instance connects to Mann–Whitney U-statistic introduced by Frank Wilcoxon and Henry Mann; computation often uses rank-sum algorithms developed in biostatistics at Johns Hopkins University and University of Oxford. Asymptotic variance and confidence intervals for AUC derive from U-statistics theory advanced by Wolfgang Hoeffding and are implemented in software libraries originating from work at Bell Labs and AT&T Laboratories.

Applications in statistics and machine learning

ROC AUC is a standard performance metric in supervised learning evaluations reported in research from Google Research, Microsoft Research, Facebook AI Research, and academic groups at Carnegie Mellon University and University of Toronto. It is used in medical diagnostics studies at Mayo Clinic and Cleveland Clinic for biomarker assessment, in information retrieval experiments influenced by Gerard Salton and Christopher Manning, and in credit scoring systems developed by practitioners from Goldman Sachs and JPMorgan Chase. AUC informs model selection alongside precision, recall, and F1 measures popularized in machine learning curricula at Coursera and edX platforms; it also appears in papers from conferences such as NeurIPS, ICML, and KDD where researchers from DeepMind and OpenAI evaluate classifiers.

Applications in pharmacokinetics

Pharmacokinetic AUC quantifies systemic exposure to a compound after administration and appears in clinical pharmacology textbooks used at University College London and Columbia University. Regulatory bioequivalence criteria by the Food and Drug Administration and the European Medicines Agency rely on comparisons of AUC and peak concentration (Cmax) when assessing generic formulations from manufacturers like Pfizer and Novartis. Noncompartmental analysis techniques using trapezoidal integration are described in guidance from World Health Organization and employed in trials at institutions such as Mayo Clinic and Massachusetts General Hospital. Model-based AUC estimation integrates compartmental parameters whose identifiability is treated in system dynamics literature influenced by Luenberger and applied in pharmacometrics groups at Certara.

Related metrics and extensions

Related summary measures include area under the precision-recall curve used in information retrieval research by Hector Zaragoza and collaborators, concordance index (c-index) applied in survival analysis as used in oncology studies at MD Anderson Cancer Center, and partial AUC variants introduced in biostatistics to emphasize clinically relevant regions, techniques developed in work at Fred Hutchinson Cancer Research Center and Dana-Farber Cancer Institute. Time-dependent AUC extensions for censored survival data draw on methods from Jerome Friedman and Trevor Hastie; multi-class adaptations for ROC surfaces have been pursued in theoretical work at ETH Zurich and Princeton University.

Historical development and notable uses

The mathematical basis of area integration emerged from the calculus of Isaac Newton and Gottfried Wilhelm Leibniz and matured through rigorous formulations by Bernhard Riemann and Henri Lebesgue. The ROC concept evolved in twentieth-century signal detection theory with contributions from David Green and John A. Swets at Psychology Research Centers and subsequently migrated to machine learning and medical diagnostics, becoming standard in journals like The Lancet, Journal of the American Medical Association, and conferences such as RSNA. Pharmacokinetic AUC became central to drug approval processes after mid‑twentieth‑century pharmacology work at Wellcome Trust laboratories and was codified in regulatory practice by the Food and Drug Administration during the expansion of clinical pharmacology in the 1970s and 1980s.

Category:Statistical measures