Abstract Introduction Machine learning has shown promising results in quantification of CAD on Coronary CT angiography (CCTA)(1). However, its diagnostic performance including vessel diameters 1.5-2.0 mm has not been adequately validated. The highly significant lower total coronary vessel volume in women versus men has been shown previously to be independent of normalization to height, weight, BMI, BSA, or left ventricular mass(2). Coronary lumen diameter less than 2 mm results in overestimation of lumen diameter by full width half maximum techniques commonly employed(2). AI-based CCTA (AI-QCT) detection and grading was performed using FDA-approved analysis, and its diagnostic performance in men versus women was compared to invasive quantitative coronary angiography (QCA). Methods Single center, retrospective evaluation of 73 clinically referred non-revascularized stable patients with 256 slice CCTA and ICA within 90 days. 34% female, mean age 64 years, BMI 28, total CAC score 691, with 24.3% obese. QCA software measurements of diameter and stenosis were obtained by a senior interventional cardiologist blinded to original AI-QCT as well as ICA/CCTA readings. AI-QCT coronary segmentation, lumen and vessel wall determination, plaque quantification and characterization, and stenosis determination was blinded to QCA/CTA/ICA interpretations. None of the 219 territories or vessels >1.5 mm as measured by the vendor were excluded. Primary endpoints were ≥50% stenosis and ≥70% stenosis per coronary territory, and per patient. Secondary endpoints were plaque features associated with diagnostic performance. Results Of a total of 73 patients, 24 F/49 M. Mean BMI 27.3/28.7, CAC 390/821 (P<0.05), noncalcified plaque volume 102/278 cu mm (p<0.00001), calcified plaque volume 159/310 cu mm (P =0.05), total coronary vessel volume 2441/3450 cu mm (P<0.0000001). At the per-vessel territory level, for the detection of >50% stenosis, specificity, sensitivity, PPV, NPV, accuracy, AUC for women: 0.84, 0.62, 0.62, 0.84, 0.78, and 0.74. For men, 0.70, 0.71, 0.52, 0.84, 0.70, and 0.81. For the detection of greater than 70% stenosis, for women, 0.92, 0.50, 0.50, 0.92, 0.86, and 0.65. For men, 0.90, 0.70, 0.52, 0.95, 0.87, and 0.88. Chi2 contingency analysis for 50% is p=4.562e^-8 and 70% p=3.73e^-11. Conclusion Despite significantly lower total calcium scores, lower noncalcified plaque volume and percent atheroma volume, and lower calcified plaque volume and percent atheroma volume, overall quantitative performance of AI-QCT as reflected by the ROC analysis is worse for women than men, at both the greater than 50% stenosis threshold, and greater than 70% stenosis threshold. The limited spatial resolution of today's conventional energy integrating polychromatic CT scanners may account for the observed lower performance for women versus men.Gender Subgroup ROC: Territory Level
N-of-1 or single subject clinical trials consider an individual patient as the sole unit of observation in a study investigating the efficacy or side-effect profiles of different interventions. The ultimate goal of an n-of-1 trial is to determine the optimal or best intervention for an individual patient using objective data-driven criteria. Such trials can leverage study design and statistical techniques associated with standard population-based clinical trials, including randomization, washout and crossover periods, as well as placebo controls. Despite their obvious appeal and wide use in educational settings, n-of-1 trials have been used sparingly in medical and general clinical settings. We briefly review the history, motivation and design of n-of-1 trials and emphasize the great utility of modern wireless medical monitoring devices in their execution. We ultimately argue that n-of-1 trials demand serious attention among the health research and clinical care communities given the contemporary focus on individualized medicine.
