Kernel‐based methods for the smooth, non‐parametric estimation of the hazard function have received considerable attention in the statistical literature. Although the mathematical properties of the kernel‐based hazard estimators have been carefully studied, their statistical properties have not. We reviewed various kernel‐based methods for hazard function estimation from right‐censored data and compared the statistical properties of these estimators through computer simulations. Our simulations covered seven distributions, three levels of random censoring, four types of bandwidth functions, two sample sizes and three types of boundary correction. We conducted a total of 504 simulation experiments with 500 independent samples each. Our results confirmed the advantages of two recent innovations in kernel estimation – boundary correction and locally optimal bandwidths. The median relative improvement (decrease) in mean square error over fixed‐bandwidth estimators without boundary correction was 3 per cent for fixed‐bandwidth estimators with left boundary correction, 52 per cent locally optimal bandwidths without boundary correction, and 66 per cent for locally optimal bandwidths with left boundary correction. The locally optimal bandwidth estimators with left boundary correction also outperformed three previously published and publicly available algorithms, with median relative improvements in mean square error of 31 per cent, 77 per cent and 80 per cent. Copyright © 1999 John Wiley & Sons, Ltd.

Large phase II trials of fluorescence and reflectance spectroscopy using a fiber optic probe in the screening and diagnostic settings for detecting cervical neoplasia have been conducted. We present accrual and histopathology data, instrumentation, data processing, and the preliminary results of interdevice consistencies throughout the progression of a trial. Patients were recruited for either a screening trial (no history of abnormal Papanicolaou smears) or a diagnostic trial (a history of abnormal Papanicolaou smears). Colposcopy identified normal and abnormal squamous, columnar, and transformation zone areas that were subsequently measured with the fiber probe and biopsied. In the course of the clinical trial, two generations of spectrometers (FastEEM2 and FastEEM3) were designed and utilized as optical instrumentation for in vivo spectroscopic fluorescence and reflectance measurements. Data processing of fluorescence and reflectance data is explained in detail and a preliminary analysis of the variability across each device and probe combination is explored. One thousand patients were recruited in the screening trial and 850 patients were recruited in the diagnostic trial. Three clinical sites attracted a diverse range of patients of different ages, ethnicities, and menopausal status. The fully processed results clearly show that consistencies exist across all device and probe combinations throughout the diagnostic trial. Based on the stratification of the data, the results also show identifiable differences in mean intensity between normal and high-grade tissue diagnosis, pre- and postmenopausal status, and squamous and columnar tissue type. The mean intensity values of stratified data show consistent separation across each of the device and probe combinations. By analyzing trial spectra, we provide more evidence that biographical variables such as menopausal status as well as tissue type and diagnosis significantly affect the data. Understanding these effects will lead to better modeling parameters when analyzing the performance of fluorescence and reflectance spectroscopy.

In the context of clinical trials, calibration protocols for optical instruments that ensure measurement accuracy and the ability to carry out meaningful comparisons of data acquired from multiple instruments are required. A series of calibration standards and procedures are presented to assess technical feasibility of optical devices for cervical precancer detection. Measurements of positive and negative standards, and tissue are made with two generations of research grade spectrometers. Calibration accuracy, ability of standards to correct and account for changes in experimental conditions, and device components are analyzed. The relative frequency of measured calibration standards is investigated retrospectively using statistical analysis of trends in instrument performance. Fluorescence measurements of standards and tissue made with completely different spectrometers show good agreement in intensity and lineshape. Frequency of wavelength calibration standards is increased to every 2 h to compensate for thermal drifts in grating mount. Variations in illumination energy detected between standards and patient measurements require probe redesign to allow for simultaneous acquisition of illumination power with every patient measurement. The use of frequent and well-characterized standards enables meaningful comparison of data from multiple devices and unambiguous interpretation of experiments among the biomedical optics community.

We report on a study designed to assess variability among three different fluorescence spectroscopy devices, four fiber optic probes, and three sets of optical calibration standards to better understand the reproducibility of measurements and interdevice comparisons of fluorescence spectroscopic data intended for clinical diagnostic use. Multiple measurements are acquired from all sets of standards using each combination of spectrometer, fiber optic probe, and optical standard. Data are processed using standard calibration methods to remove instrument-dependant responses. Processed spectra are analyzed using an analysis of variance to assess the percent variance explained by each factor that was statistically significant. Analysis of processed data confirms statistically significant differences among the spectrometers and fiber optic probes. However, no differences are found when varying calibration standards or measurement date and time. The spectrometers and fiber optic probes are significant sources of variability, but appropriate data processing substantially reduces these effects. Studies of inter- and intradevice variability are important methodological issues for optical device trials and must be included in the quality assurance studies for the clinical trial design.

No abstract