SummaryThe Youden Index is a frequently used summary measure of the ROC (Receiver Operating Characteristic) curve. It both, measures the effectiveness of a diagnostic marker and enables the selection of an optimal threshold value (cutoff point) for the marker. In this paper we compare several estimation procedures for the Youden Index and its associated cutoff point. These are based on (1) normal assumptions; (2) transformations to normality; (3) the empirical distribution function; (4) kernel smoothing. These are compared in terms of bias and root mean square error in a large variety of scenarios by means of an extensive simulation study. We find that the empirical method which is the most commonly used has the overall worst performance. In the estimation of the Youden Index the kernel is generally the best unless the data can be well transformed to achieve normality whereas in estimation of the optimal threshold value results are more variable.

Neural networks have received considerable attention recently, mostly by non-statisticians. They are considered by many to be very promising tools for classification and prediction. In this paper we present an approach to modelling censored survival data using the input-output relationship associated with a simple feed-forward neural network as the basis for a non-linear proportional hazards model. This approach can be extended to other models used with censored survival data. The proportional hazards neural network parameters are estimated using the method of maximum likelihood. These maximum likelihood based models can be compared, using readily available techniques such as the likelihood ratio test and the Akaike criterion. The neural network models are illustrated using data on the survival of men with prostatic carcinoma. A method of interpreting the neural network predictions based on the factorial contrasts is presented.

The area under the receiver operating characteristic curve is frequently used as a measure for the effectiveness of diagnostic markers. In this paper we discuss and compare estimation procedures for this area. These are based on (i) the Mann-Whitney statistic; (ii) kernel smoothing; (iii) normal assumptions; (iv) empirical transformations to normality. These are compared in terms of bias and root mean square error in a large variety of situations by means of an extensive simulation study. Overall we find that transforming to normality usually is to be preferred except for bimodal cases where kernel methods can be effective.

Background Our objective was to determine whether preconception-initiated low dose aspirin (LDA) improved live birth rates in women with one to two prior pregnancy losses. Methods This multi-center, block-randomized, double-blind, placebo-controlled trial recruited from four medical centers in the US (2006–2012). Women aged 18–40 years attempting pregnancy were stratified by eligibility criteria: “original”: women with one loss <20 weeks’ gestation during the past year; or “expanded”: women with one to two prior losses regardless of gestational length or time of loss. Women were block-randomized (615 LDA, 613 placebo) by center and eligibility stratum. Preconception-initiated daily LDA (81 mg/day) was compared with placebo for up to six menstrual cycles; for those who conceived, study treatment continued until 36 weeks’ gestation. The primary outcome was live birth rate. The trial was registered on ClinicalTrials.gov (#NCT00467363). Findings Overall, 1078 women completed the trial (LDA 535, placebo 543). Live birth rates were 58% (309/535) in women assigned LDA vs. 53% placebo (286/543; risk difference [RD] 5%; 95% confidence interval [CI] −0·8, 11). Pregnancy loss rates were similar between groups (13% [68/535] LDA, 12% [65/543] placebo; p=0·7812). In the original stratum, live birth rates were 62% (151/242) LDA vs. 53% (133/250) placebo (RD 9%; 95% CI 0·5, 18), and in the expanded, 54% (158/293) LDA vs. 52% (153/293) placebo (RD 2%; 95% CI −6, 10). Major adverse events were similar between treatment arms. LDA was associated with increased bleeding per vaginam, but this was not associated with losses. Interpretation Preconception-initiated LDA was not significantly associated with live birth or pregnancy loss among women with one to two prior losses. However, higher live birth rates were observed among women with a single documented loss at <20 weeks’ gestation during the previous year. LDA is not recommended for the prevention of pregnancy loss.

The area under the receiver operating characteristic curve is the most commonly used measure of the ability of a biomarker to distinguish between two populations. Some markers are subject to substantial measurement error. Under normality assumptions, the authors develop a confidence interval procedure for the area under the receiver operating characteristic curve that adjusts for measurement error. This procedure assumes the availability of data from a reliability study of the biomarker. A simulation study was used to check the validity of the proposed confidence interval. Furthermore, it was shown that not adjusting for measurement error could result in a serious understatement of the effectiveness of the biomarker.