Glaucoma is the primary cause of vision loss in Japan. The most important glaucoma therapy is to decrease intraocular pressure (IOP) for preventing visual field defects in the pre-stage of vision loss. Considering a systematic diurnal variation of IOP, Kuwayama et al. (2006) proposed to use a circular linear mixed effect (CLME) model for evaluating the efficacy of therapy on IOP decrease for patients with normal tension glaucoma (NTG) and applied it to the data analysis in a clinical trial (Nipradilol trial) with 28 NTG patients. In this application, there occurred an issue that the parameter estimates were different depending on the method of estimation and the best method was not identified. We, therefore, compared six methods for parameter estimation (standard two-stage (STS) method, global two-stage (GTS) method, first order approximation (FOA) method, Laplacian approximation (LAP) method, Monte Carlo integration (MCI) method and Gaussian quadrature (GAUS) method) through a simulation experiment with the bias and square root of mean squared error as the criteria for evaluation. The GAUS method proved to be superior to others in realizing least bias and mean squared error under various simulation conditions, although it was most time consuming.
Meta-analysis of randomized controlled trials is a widely used study methodology and it is considered to provide the highest level of evidence. The results of such analysis, however, by the nature of this methodology, may be affected by a very serious bias referred to as publication bias. Although the trim-and-fill method has been proposed as a means of adjusting for publication bias, it does not necessarily provide a suitable correction under realistic circumstances. This article proposes a new method to correct for publication bias based on p-value and evaluates the performance of this method by means of simulations. It is shown that the performance of the proposed method is superior to that of the trim-and-fill method under realistic situations.
Michiels et al. (2005) showed that a list of genes identified as predictors of prognosis via a non-repeated training — validation approach is unstable and advocate the validation by repeated random sampling. They considered that the genes which were selected as top 50 genes in more than half of their jackknife samples were stable for prediction. However, there is no rationale of the determination of the length of the gene list and the threshold of stability. Since evaluating an accumulation of low p-values in the repeated random sampling is essentially required for a stability assessment, it is better to compare the distribution of p-values of a gene observed with the distribution of p-values under the null hypothesis directly. In this study, the Quantile-Quantile plot (Q-Q plot) of p-values with null reference was proposed for this purpose. We applied the proposed method to a clinical data for primary breast cancer. The Q-Q plot approach can reveal that the genes with a similar p-value in the ordinary analysis have different p-value distributions in the repeated random sampling, and the gene with low p-values accumulated in the repeated random sampling could be evaluated according to the reference lines in the Q-Q plot.
In epidemiological studies there are numerous challenges to finding an appropriate experimental design for studying the relationship between an exposure and a disease. A number of designs for observational studies have been investigated; the cohort study and the standard case-control study are the primary types. The nested case-control study design combines features of both. In the nested case-control study some information is available for all cohort members and additional information is obtained using a case-control design based on incidence density sampling rather than cumulative prevalence sampling. To get higher efficiency, it is important to effectively use the information already available on cohort members when we choose the controls. The counter-matched design is one such method, which chooses the controls based on exposure status or a surrogate measure of the exposure. It has high efficiency to estimate the effect of the exposure or interaction between the exposure and another risk factor. Unfortunately, there are few examples of application of this design to real situations. To encourage use of the counter-matched design, we describe the basic concept and statistical methods underlying the method. Some examples are presented to illustrate how to implement the design. We also propose some extensions to provide a more flexible sampling design.