Central statistical monitoring can both optimize on-site monitoring and improve data quality and as such provides a cost-effective way of meeting regulatory requirements for clinical trials.

Multicenter studies are widely used to meet accrual targets in clinical trials. Clinical data monitoring is required to ensure the quality and validity of the data gathered across centers. One approach to this end is central statistical monitoring, which aims at detecting atypical patterns in the data by means of statistical methods. In this context, we consider the simple case of a continuous variable, and we propose a detection procedure based on a linear mixed-effects model to detect location differences between each center and all other centers. We describe the performance of the procedure as a function of contamination rate and signal-to-noise ratio. We investigate the effect of center size and variance structure and illustrate the use of the procedure using data from two multicenter clinical trials.

Summary Liver transplantation (LT) is a validated treatment for selected cirrhotics with hepatocellular cancer (HCC). A retrospective single center study including 137 recipients having proven HCC was done to refine inclusion criteria for LT as well as to look at impact of locoregional treatment (LRT) on outcome. At pre‐LT imaging, 42 (30.6%) patients were Milan criteria (MC)‐OUT; 28 (20.4%) were University of California San Francisco criteria (UCSFC)‐OUT. Pre‐LT LRT was performed in 109 (79.6%) patients. Multivariate analysis identified four factors to be independently predictive of recurrence: tumour number >3, AFP level ≥400 ng/ml, microvascular invasion and rejection needing anti‐lymphocytic antibodies. When considering pre‐transplant variables only, AFP level ≥400 ng/ml (HR = 5.13; P < 0.0001) was the unique risk factor for recurrence; conversely, application of LRT was protective (HR = 0.42; P = 0.04). MC‐IN patients having LRT (n = 79) had the best 5‐year tumour‐free survival (TFS) (91.6%). MC‐IN patients without LRT (n = 16) and MC‐OUT patients with LRT (n = 30) had similar good TFS (72.7% vs.77.5%); finally MC‐OUT patients without LRT (n = 12) had the worst results (45.0%; vs. 1st group: P < 0.0001). Immediate pre‐LT AFP and aggressive pre‐transplant LRT strategy, especially in MC‐OUT patients, are both important elements to further expand inclusion criteria without compromising long‐term results of HCC liver recipients.

As part of central statistical monitoring of multicenter clinical trial data, we propose a procedure based on the beta-binomial distribution for the detection of centers with atypical values for the probability of some event. The procedure makes no assumptions about the typical event proportion and uses the event counts from all centers to derive a reference model. The procedure is shown through simulations to have high sensitivity and high specificity if the contamination rate is small and the atypical event proportions are the result of some systematic shift in the underlying data generating mechanism.

Background/Aims A risk-based approach to clinical research may include a central statistical assessment of data quality. We investigated the operating characteristics of unsupervised statistical monitoring aimed at detecting atypical data in multicenter experiments. The approach is premised on the assumption that, save for random fluctuations and natural variations, data coming from all centers should be comparable and statistically consistent. Unsupervised statistical monitoring consists of performing as many statistical tests as possible on all trial data, in order to detect centers whose data are inconsistent with data from other centers. Methods We conducted simulations using data from a large multicenter trial conducted in Japan for patients with advanced gastric cancer. The actual trial data were contaminated in computer simulations for varying percentages of centers, percentages of patients modified within each center and numbers and types of modified variables. The unsupervised statistical monitoring software was run by a blinded team on the contaminated data sets, with the purpose of detecting the centers with contaminated data. The operating characteristics (sensitivity, specificity and Youden’s J-index) were calculated for three detection methods: one using the p-values of individual statistical tests after adjustment for multiplicity, one using a summary of all p-values for a given center, called the Data Inconsistency Score, and one using both of these methods. Results The operating characteristics of the three methods were satisfactory in situations of data contamination likely to occur in practice, specifically when a single or a few centers were contaminated. As expected, the sensitivity increased for increasing proportions of patients and increasing numbers of variables contaminated. The three methods showed a specificity better than 93% in all scenarios of contamination. The method based on the Data Inconsistency Score and individual p-values adjusted for multiplicity generally had slightly higher sensitivity at the expense of a slightly lower specificity. Conclusions The use of brute force (a computer-intensive approach that generates large numbers of statistical tests) is an effective way to check data quality in multicenter clinical trials. It can provide a cost-effective complement to other data-management and monitoring techniques.