Current issues in the design and interpretation of clinical trials.

Though there have been considerable improvements in the use of statistical methods for clinical trials in recent years, there remain major practical difficulties in the design and interpretation of many trials. This paper concentrates on problems relating to randomisation, the overemphasis on significance testing, and the inadequate size of many trials. Each topic is illustrated by examples from recent trials.     

Best practices for design and implementation of human clinical trials studying dietary oils

Dietary oils are a significant contributor to overall energy and fatty acid intakes. Changes in the amount and/or type of dietary oils consumed have the potential to impact human health. Clinical trials represent the gold standard for testing the health impacts of such changes in dietary oils. The objective of this review is to explore best practices for clinical trials examining impacts of dietary oils including 1) pre-clinical topics such as research question generation, study design, participant population, outcome measures and intervention product selection and/or preparation; 2) clinical trial implementation topics such as recruitment, trial management, record keeping and compliance monitoring; and 3) post-clinical trial topics dealing with sample analysis and storage as well as management, publication and data access. The use of digital case report forms, and the best practices in reporting and publishing results are also addressed. In summary, properly designed and implemented clinical trials studying dietary oils produce strong scientific evidence-guiding their use.     

Multiplicity of inferences in clinical trials: adjustment methods, clinical interpretation issues

Multiplicity of inferences is present in a large majority of clinical trials and conducts to false analyses or interpretation issues. The main risk consists in false positive conclusions. A large number of statistical methods is available for controlling the rate of false positive conclusions. But formal adjustment is not necessary in all cases and depends on the aims of the study.     

Expression and purification of cGMP grade NY-ESO-1 for clinical trials

NY-ESO-1 is a cancer testis antigen expressed in numerous cancers. Initial tests have shown its efficacy as a cancer vaccine, stimulating the body's own immune response against the invading tumor. To produce enough material for phase I clinical trials, a process using current good manufacturing practices to produce clinical grade material was developed and executed. His-tagged NY-ESO-1 was expressed in C41DE3 Escherichia coli under control of the T-7 promoter. NY-ESO-1 was produced in a 20 L fed-batch fermentation utilizing a pH-stat control scheme. The protein was then purified from inclusion bodies using a three-column process that achieved a yield of over 3.4 g and endotoxin below the detection limit of 0.005 EU/μg protein.     

Stem cell therapies in clinical trials: workshop on best practices and the need for harmonization

A workshop addressing regulation of clinical implementation of stem cell therapies preceded the ISSCR 8th Annual Meeting, cosponsored by the International Society for Stem Cell Research, the California Institute for Regenerative Medicine and the International Society for Cellular Therapy.     

Computer system for assisting with clinical interpretation of tumour marker data.

OBJECTIVE--To design and evaluate a computer advisory system for the treatment of gestational trophoblastic tumour. DESIGN--A comparison of clinicians'' treatment decisions with those of the computer system. Two datasets were used: one to calibrate the system and one to independently evaluate it. SETTING--Department of medical oncology. PATIENTS--Computerised records of 290 patients with low risk gestational trophoblastic tumour for whom the advisory system could predict the adequacy of treatment. The calibration set comprised patients admitted during 1979-86(227) and the test set patients during 1986-89(63). MAIN OUTCOME MEASURES--The system''s accuracy in predicting need to change treatment compared with clinicians'' actions. The mean time faster that the system was in predicting the need to change treatment. RESULTS--On the calibration dataset the system was 94% (164/174) accurate in predicting patients whose treatment was adequate, recommending change when none occurred in only 10 (6%) patients. In patients whose treatment was changed the system recommended change earlier than clinicians in 39/53 cases (74%), with a mean time advantage of 14.9 (SE 2.02) days. On the test dataset the system had an accuracy of 91% (31/34) in predicting treatment adequacy and a false positive rate of 9% (3/34). The system recommended change earlier than clinicians in 22/29 cases (76%), with a mean time advantage of 12.5 (2.22) days. CONCLUSIONS--The computer advisory system could improve patient management by reducing the time spent receiving ineffective treatment. This has implications for both patient time and clinical costs.     

Blinded and unblinded internal pilot study designs for clinical trials with count data

Internal pilot studies are a popular design feature to address uncertainties in the sample size calculations caused by vague information on nuisance parameters. Despite their popularity, only very recently blinded sample size reestimation procedures for trials with count data were proposed and their properties systematically investigated. Although blinded procedures are favored by regulatory authorities, practical application is somewhat limited by fears that blinded procedures are prone to bias if the treatment effect was misspecified in the planning. Here, we compare unblinded and blinded procedures with respect to bias, error rates, and sample size distribution. We find that both procedures maintain the desired power and that the unblinded procedure is slightly liberal whereas the actual significance level of the blinded procedure is close to the nominal level. Furthermore, we show that in situations where uncertainty about the assumed treatment effect exists, the blinded estimator of the control event rate is biased in contrast to the unblinded estimator, which results in differences in mean sample sizes in favor of the unblinded procedure. However, these differences are rather small compared to the deviations of the mean sample sizes from the sample size required to detect the true, but unknown effect. We demonstrate that the variation of the sample size resulting from the blinded procedure is in many practically relevant situations considerably smaller than the one of the unblinded procedures. The methods are extended to overdispersed counts using a quasi‐likelihood approach and are illustrated by trials in relapsing multiple sclerosis.     

On the choice of times for data analysis in group sequential clinical trials

Planned interim analysis of randomized clinical trials has been implemented for over a decade. While the initial proposal advocated analyzing after equal numbers of patients were evaluated, a later modification by Lan and DeMets (1983, Biometrika 70, 659-663) allowed for more flexible boundaries. Rather than fixing the times of analysis at equal numbers of patients, they fixed the rate at which overall alpha was used up according to a use function alpha * (t) on t in with alpha * (0) = 0 and alpha * (1) = alpha. Here we consider how flexible Lan and DeMets' procedure is. We show that the choice of alpha * (t) for a particular trial affects the permissible analysis times if other desirable properties of the sequence of nominal significance levels are to hold. To overcome the difficulties posed by patterns of late analysis, piecewise linear convex use functions are proposed.     

Analyzing excessive no changes in clinical trials with clustered data

This article considers clinical trials in which the efficacy measure is taken from several sites within each patient, such as the alveolar bone height of the tooth sites, or bone mineral densities of the lumbar spine sites. Since usually only a small portion of these sites will exhibit changes, the conventional method using per patient average gives a diluted result due to excessive no changes in the data. Different methods have been proposed for this type of data in the case where the observations are mutually independent. This includes the popular "two-part model" (Lachenbruch, 2001, Statistics in Medicine 20, 1215-1234; 2002, Statistical Methods in Medical Research 11, 297-302), which is related to the "composite approach" for discrete and continuous data in Shih and Quan (1997, Statistics in Medicine16, 1225-1239; 2001, Statistica Sinica 11, 53-62). In this article, we model the data with excessive zeros (no changes) in clustered data using a mixture of distributions, and taking into account possible measurement errors. This mixture model includes the two-part model as a special case when one component of the mixture degenerates.     

Problems in dealing with missing data and informative censoring in clinical trials

A common problem in clinical trials is the missing data that occurs when patients do not complete the study and drop out without further measurements. Missing data cause the usual statistical analysis of complete or all available data to be subject to bias. There are no universally applicable methods for handling missing data. We recommend the following: (1) Report reasons for dropouts and proportions for each treatment group; (2) Conduct sensitivity analyses to encompass different scenarios of assumptions and discuss consistency or discrepancy among them; (3) Pay attention to minimize the chance of dropouts at the design stage and during trial monitoring; (4) Collect post-dropout data on the primary endpoints, if at all possible; and (5) Consider the dropout event itself an important endpoint in studies with many.     

Next-generation text-mining mediated generation of chemical response-specific gene sets for interpretation of gene expression data

Background

Availability of chemical response-specific lists of genes (gene sets) for pharmacological and/or toxic effect prediction for compounds is limited. We hypothesize that more gene sets can be created by next-generation text mining (next-gen TM), and that these can be used with gene set analysis (GSA) methods for chemical treatment identification, for pharmacological mechanism elucidation, and for comparing compound toxicity profiles.

Methods

We created 30,211 chemical response-specific gene sets for human and mouse by next-gen TM, and derived 1,189 (human) and 588 (mouse) gene sets from the Comparative Toxicogenomics Database (CTD). We tested for significant differential expression (SDE) (false discovery rate -corrected p-values < 0.05) of the next-gen TM-derived gene sets and the CTD-derived gene sets in gene expression (GE) data sets of five chemicals (from experimental models). We tested for SDE of gene sets for six fibrates in a peroxisome proliferator-activated receptor alpha (PPARA) knock-out GE dataset and compared to results from the Connectivity Map. We tested for SDE of 319 next-gen TM-derived gene sets for environmental toxicants in three GE data sets of triazoles, and tested for SDE of 442 gene sets associated with embryonic structures. We compared the gene sets to triazole effects seen in the Whole Embryo Culture (WEC), and used principal component analysis (PCA) to discriminate triazoles from other chemicals.

Results

Next-gen TM-derived gene sets matching the chemical treatment were significantly altered in three GE data sets, and the corresponding CTD-derived gene sets were significantly altered in five GE data sets. Six next-gen TM-derived and four CTD-derived fibrate gene sets were significantly altered in the PPARA knock-out GE dataset. None of the fibrate signatures in cMap scored significant against the PPARA GE signature. 33 environmental toxicant gene sets were significantly altered in the triazole GE data sets. 21 of these toxicants had a similar toxicity pattern as the triazoles. We confirmed embryotoxic effects, and discriminated triazoles from other chemicals.

Conclusions

Gene set analysis with next-gen TM-derived chemical response-specific gene sets is a scalable method for identifying similarities in gene responses to other chemicals, from which one may infer potential mode of action and/or toxic effect.     

Improving efficiency of inference in clinical trials with external control data

Suppose we are interested in the effect of a treatment in a clinical trial. The efficiency of inference may be limited due to small sample size. However, external control data are often available from historical studies. Motivated by an application to Helicobacter pylori infection, we show how to borrow strength from such data to improve efficiency of inference in the clinical trial. Under an exchangeability assumption about the potential outcome mean, we show that the semiparametric efficiency bound for estimating the average treatment effect can be reduced by incorporating both the clinical trial data and external controls. We then derive a doubly robust and locally efficient estimator. The improvement in efficiency is prominent especially when the external control data set has a large sample size and small variability. Our method allows for a relaxed overlap assumption, and we illustrate with the case where the clinical trial only contains a treated group. We also develop doubly robust and locally efficient approaches that extrapolate the causal effect in the clinical trial to the external population and the overall population. Our results also offer a meaningful implication for trial design and data collection. We evaluate the finite-sample performance of the proposed estimators via simulation. In the Helicobacter pylori infection application, our approach shows that the combination treatment has potential efficacy advantages over the triple therapy.     

A random effects model for multivariate failure time data from multicenter clinical trials

A random effects model for analyzing multivariate failure time data is proposed. The work is motivated by the need for assessing the mean treatment effect in a multicenter clinical trial study, assuming that the centers are a random sample from an underlying population. An estimating equation for the mean hazard ratio parameter is proposed. The proposed estimator is shown to be consistent and asymptotically normally distributed. A variance estimator, based on large sample theory, is proposed. Simulation results indicate that the proposed estimator performs well in finite samples. The proposed variance estimator effectively corrects the bias of the naive variance estimator, which assumes independence of individuals within a group. The methodology is illustrated with a clinical trial data set from the Studies of Left Ventricular Dysfunction. This shows that the variability of the treatment effect is higher than found by means of simpler models.