To tolerate or to agree: A tutorial on tolerance intervals in method comparison studies with BivRegBLS R Package

Exploring the Potential of Artificial Intelligence in Pediatric Echocardiography-Preliminary Results from the First Pediatric Study Using AI Software Developed for Adults

(1) Background: Transthoracic echocardiography is the first-line non-invasive investigation for assessing pediatric patients' cardiac anatomy, physiology, and hemodynamics, based on its accessibility and portability, but complete anatomic and hemodynamic assessment is time-consuming. (2) Aim: This study aimed to determine whether an automated software developed for adults could be effectively used for the analysis of pediatric echocardiography studies without prior training. (3) Materials and Methods: The study was conducted at the University Hospital of Bordeaux between August and September 2022 and included 45 patients with normal or near normal heart architecture who underwent a 2D TTE. We performed Spearman correlation and Bland-Altman analysis. (4) Results: The mean age of our patients at the time of evaluation was 8.2 years ± 5.7, and the main reason for referral to our service was the presence of a heart murmur. Bland-Altman analysis showed good agreement between AI and the senior physician for two parameters (aortic annulus and E wave) regardless of the age of the children included in the study. A good agreement between AI and physicians was also achieved for two other features (STJ and EF) but only for patients older than 9 years. For other features, either a good agreement was found between physicians but not with the AI, or a poor agreement was established. In the first case, maybe proper training of the AI could improve the measurement, but in the latter case, for now, it seems unrealistic to expect to reach a satisfactory accuracy. (5) Conclusion: Based on this preliminary study on a small cohort group of pediatric patients, the AI soft originally developed for the adult population, had provided promising results in the evaluation of aortic annulus, STJ, and E wave.

Determining reference ranges and sample sizes in parallel-group studies

BACKGROUND AND OBJECTIVES

Reference ranges are widely used to locate the major range of the target probability distribution. When future measurements fall outside the reference range, they are classified as atypical and require further investigation. The fundamental principles and statistical properties of reference ranges are closely related to those of tolerance interval procedures. Existing investigations of reference ranges and tolerance intervals mainly devoted to the primitive cases of one- and paired-sample designs. Although reference ranges hold considerable promise for parallel group designs, the corresponding methodological and computational issues for determining reference limits and sample sizes have not been adequately addressed.

METHODS

This paper describes a complete collection of one- and two-sided reference ranges for assessing measurement differences in parallel-group studies that assume variance homogeneity.

RESULTS

The problem of sample size determination for precise reference ranges is also examined under the expected half-width and assurance probability considerations. Unlike the current methods, the suggested sample size criteria explicitly accommodate desired interval width in precise interval estimation.

CONCLUSIONS

Theoretical examinations and empirical assessments are presented to validate the usefulness of the proposed reference range and sample size procedures. To enhance the usages of the recommended techniques in practical applications, computer programs are developed for efficient calculation and exact analysis. A real data example regarding tablet absorption rate and extent is presented to illustrate the suggested assessments between two drug formulations.

Continuous Glucose Deviation Interval and Variability Analysis (CG-DIVA): A Novel Approach for the Statistical Accuracy Assessment of Continuous Glucose Monitoring Systems

BACKGROUND

The accuracy of continuous glucose monitoring (CGM) systems is crucial for the management of glucose levels in individuals with diabetes mellitus. However, the discussion of CGM accuracy is challenged by an abundance of parameters and assessment methods. The aim of this article is to introduce the Continuous Glucose Deviation Interval and Variability Analysis (CG-DIVA), a new approach for a comprehensive characterization of CGM point accuracy which is based on the U.S. Food and Drug Administration requirements for "integrated" CGM systems.

METHODS

The statistical concept of tolerance intervals and data from two approved CGM systems was used to illustrate the CG-DIVA.

RESULTS

The CG-DIVA characterizes the expected range of deviations of the CGM system from a comparison method in different glucose concentration ranges and the variability of accuracy within and between sensors. The results of the CG-DIVA are visualized in an intuitive and straightforward graphical presentation. Compared with conventional accuracy characterizations, the CG-DIVA infers the expected accuracy of a CGM system and highlights important differences between CGM systems. Furthermore, it provides information on the incidence of large errors which are of particular clinical relevance. A software implementation of the CG-DIVA is freely available (https://github.com/IfDTUlm/CGM_Performance_Assessment).

CONCLUSIONS

We argue that the CG-DIVA can simplify the discussion and comparison of CGM accuracy and could replace the high number of conventional approaches. Future adaptations of the approach could thus become a putative standard for the accuracy characterization of CGM systems and serve as the basis for the definition of future CGM performance requirements.

Validation of 3-dimensional mathematical computation of feline bladder volume using orthogonal ultrasonographic bladder images

Urine output measurement is critical for the management of hospitalized cats and their underlying conditions. Ultrasound-guided estimation of urinary bladder volume (UBV) is a non-invasive surrogate measurement that can provide important clinical information. The purpose of this study was to validate the accuracy of a novel 3D computation method in estimating UBV using 2D point-of-care ultrasonographic images. Bladder volume estimation was performed using coordinates from bladder circumference tracings on paired longitudinal and transverse ultrasonographic images (n = 359) aligned in 3D space for mathematical algorithmic computation. Ultrasonographic images were obtained by 2 different observers at 18 different time points on 10 healthy, purpose-bred male cats under general anesthesia in sternal recumbency. Actual urine volumes were measured via urinary catheterization and compared to UBV estimations using Lin's concordance correlation coefficient and Bland-Altman analysis. Estimation of UBV using the 3D computational bladder circumference tracing method showed moderate strength-of-agreement with actual bladder volume (ρ_c = 0.94 to 0.95) with clinically insignificant bias (3D computation-derived minus actual volume) of -1.96 mL (IQR = -3.89 to -0.57 mL, P < 0.001) and -2.42 mL (IQR = -4.64 to -0.66 mL, P < 0.001) for the 2 observers, respectively. Our study demonstrated acceptable accuracy of 3D computation method for UBV estimation in healthy cats. This method may provide a bridging alternative until 3D ultrasound becomes more readily accessible.

Assessment of Operator Reliability in Ultrasound of the Median and Ulnar Nerve Using Bland-Altman Analysis

Currently, there is no standardized method to evaluate operator reliability in nerve ultrasound. A short prospective protocol using Bland–Altman analysis was developed to assess the level of agreement between operators with different expertise levels. A control rater without experience in nerve ultrasound, three novices after two months of training, an experienced rater with two years of experience, and a reference rater performed blinded ultrasound examinations of the left median and ulnar nerve in 42 nerve sites in healthy volunteers. The precision of Bland–Altman agreement analysis was tested using the Preiss–Fisher procedure. Intraclass correlation coefficients (ICC), coefficients of variation, and Bland–Altman limits of agreement were calculated. The sample size calculation and Preiss–Fisher procedure showed a sufficient precision of Bland–Altman agreement analysis. Limits of agreement of all trained novices ranged from 2.0 to 2.9 mm² and were within the test’s maximum tolerated difference. Ninety-five percent confidence intervals of limits of agreement revealed a higher precision in the experienced rater’s measurements. Operator reliability in nerve ultrasound of the median and ulnar nerve arm nerves can be evaluated with a short prospective controlled protocol using Bland–Altman statistics, allowing a clear distinction between an untrained rater, trained novices after two months of training, and an experienced rater.