When is a seamless study desirable? Case studies from different pharmaceutical sponsors.

BACKGROUND: Inferentially seamless studies are one of the best-known adaptive trial designs. Statistical inference for these studies is a well-studied problem. Regulatory guidance suggests that statistical issues associated with study conduct are not as well understood. Some of these issues are caused by the need for early pre-specification of the phase III design and the absence of sponsor access to unblinded data. Before statisticians decide to choose a seamless IIb/III design for their programme, they should consider whether these pitfalls will be an issue for their programme. METHODS: We consider four case studies. Each design met with varying degrees of success. We explore the reasons for this variation to identify characteristics of drug development programmes that lend themselves well to inferentially seamless trials and other characteristics that warn of difficulties. RESULTS: Seamless studies require increased upfront investment and planning to enable the phase III design to be specified at the outset of phase II. Pivotal, inferentially seamless studies are unlikely to allow meaningful sponsor access to unblinded data before study completion. This limits a sponsor's ability to reflect new information in the phase III portion. CONCLUSIONS: When few clinical data have been gathered about a drug, phase II data will answer many unresolved questions. Committing to phase III plans and study designs before phase II begins introduces extra risk to drug development. However, seamless pivotal studies may be an attractive option when the clinical setting and development programme allow, for example, when revisiting dose selection.

The statistician's role in the prevention of missing data.

Considerable statistical research has been performed in recent years to develop sophisticated statistical methods for handling missing data and dropouts in the analysis of clinical trial data. However, if statisticians and other study team members proactively set out at the trial initiation stage to assess the impact of missing data and investigate ways to reduce dropouts, there is considerable potential to improve the clarity and quality of trial results and also increase efficiency. This paper presents a Human Immunodeficiency Virus (HIV) case study where statisticians led a project to reduce dropouts. The first step was to perform a pooled analysis of past HIV trials investigating which patient subgroups are more likely to drop out. The second step was to educate internal and external trial staff at all levels about the patient types more likely to dropout, and the impact this has on data quality and sample sizes required. The final step was to work collaboratively with clinical trial teams to create proactive plans regarding focused retention efforts, identifying ways to increase retention particularly in patients most at risk. It is acknowledged that identifying the specific impact of new patient retention efforts/tools is difficult because patient retention can be influenced by overall study design, investigational product tolerability profile, current standard of care and treatment access for the disease under study, which may vary over time. However, the implementation of new retention strategies and efforts within clinical trial teams attests to the influence of the analyses described in this case study.

The inclusion of historical control data may reduce the power of a confirmatory study.

Large sample sizes in clinical trials increase the cost of clinical research and delay the availability of new treatments. Fewer patients could be recruited into clinical trials if historical data on the comparator could be used reliably in a trial's analysis. However, old trials may bias rather than augment data from a new trial if, for example, the standard of care has improved over time. A hierarchical model for the data from the current and historical trials decreases the weight given to the historical data in line with the discrepancy between the results from the different trials. This reduces the risk of substantial bias. This paper shows that this down-weighting is not sufficiently sensitive to differences in the response rates between trials. Motivated by recent trials in HIV, this paper proposes and examines a more conservative weighting of historical data. Simulation showed that both the standard hierarchical and the proposed weighting of historical data led to Type II error rates worse than those attained by ignoring the historical data completely. This underlines the risks of including historical data in the primary analysis of a trial for which strict control of error rates is paramount.

Adjusting for covariates in non-inferiority studies with margins defined as risk differences.

Adjusting for covariates makes efficient use of data and can improve the precision of study results or even reduce sample sizes. There is no easy way to adjust for covariates in a non-inferiority study for which the margin is defined as a risk difference. Adjustment is straightforward on the logit scale, but reviews of clinical studies suggest that the analysis is more often conducted on the more interpretable risk-difference scale. We examined four methods that allow for adjustment on the risk-difference scale: stratified analysis with Cochran-Mantel-Haenszel (CMH) weights, binomial regression with an identity link, the use of a Taylor approximation to convert results from the logit to the risk-difference scale and converting the risk-difference margin to the odds-ratio scale. These methods were compared using simulated data based on trials in HIV. We found that the CMH had the best trade-off between increased efficiency in the presence of predictive covariates and problems in analysis at extreme response rates. These results were shared with regulatory agencies in Europe and the USA, and the advice received is described.

Informing the selection of futility stopping thresholds: case study from a late-phase clinical trial.

In an environment where (i) potential risks to subjects participating in clinical studies need to be managed carefully, (ii) trial costs are increasing, and (iii) there are limited research resources available, it is necessary to prioritize research projects and sometimes re-prioritize if early indications suggest that a trial has low probability of success. Futility designs allow this re-prioritization to take place. This paper reviews a number of possible futility methods available and presents a case study from a late-phase study of an HIV therapeutic, which utilized conditional power-based stopping thresholds. The two most challenging aspects of incorporating a futility interim analysis into a trial design are the selection of optimal stopping thresholds and the timing of the analysis, both of which require the balancing of various risks. The paper outlines a number of graphical aids that proved useful in explaining the statistical risks involved to the study team. Further, the paper outlines a decision analysis undertaken which combined expectations of drug performance with conditional power calculations in order to produce probabilities of different interim and final outcomes, and which ultimately led to the selection of the final stopping thresholds.

HIV prevention trial design in an era of effective pre-exposure prophylaxis.

Pre-exposure prophylaxis (PrEP) has demonstrated remarkable effectiveness protecting at-risk individuals from HIV-1 infection. Despite this record of effectiveness, concerns persist about the diminished protective effect observed in women compared with men and the influence of adherence and risk behaviors on effectiveness in targeted subpopulations. Furthermore, the high prophylactic efficacy of the first PrEP agent, tenofovir disoproxil fumarate/emtricitabine (TDF/FTC), presents challenges for demonstrating the efficacy of new candidates. Trials of new agents would typically require use of non-inferiority (NI) designs in which acceptable efficacy for an experimental agent is determined using pre-defined margins based on the efficacy of the proven active comparator (i.e. TDF/FTC) in placebo-controlled trials. Setting NI margins is a critical step in designing registrational studies. Under- or over-estimation of the margin can call into question the utility of the study in the registration package. The dependence on previous placebo-controlled trials introduces the same issues as external/historical controls. These issues will need to be addressed using trial design features such as re-estimated NI margins, enrichment strategies, run-in periods, crossover between study arms, and adaptive re-estimation of sample sizes. These measures and other innovations can help to ensure that new PrEP agents are made available to the public using stringent standards of evidence.

Medium-term variability of blood pressure and potential underdiagnosis of hypertension in patients with previous transient ischemic attack or minor stroke.

BACKGROUND AND PURPOSE: Blood pressure (BP) is a major risk factor for stroke. However, the variability of systolic and diastolic BP (SBP and DBP) means that single measurements do not provide a reliable measure of usual BP. Although 24-hour ambulatory BP monitoring can correct for the effects of short-term variation, there is also important medium-term variability. The extent of medium-term variability in BP is most marked in patients with a previous transient ischemic attack (TIA) or stroke. We studied the potential impact of this variability on the likely recognition of hypertension. METHODS: We analyzed multiple repeated measurements of BP in 3 large cohorts with a TIA or minor stroke: the UK-TIA trial (n=2098), the Dutch TIA trial (n=2953), and the European Carotid Surgery Trial (ECST; n=2646). Regression dilution ratios and coefficients of variation were calculated for SBP and DBP from baseline and repeated measurements during the subsequent 12 months. Categorization based on single baseline measurements was also compared with categorization based on the subsequent "usual" BP. RESULTS: The correlation between measurements of BP at baseline and 3 to 5 months later was poor (R(2) from 0.17 to 0.31 for SBP and from 0.10 to 0.20 for DBP). Categorization of patients by baseline values resulted in substantial misclassification in relation to usual BP. For example, of patients with an SBP <140 mm Hg at baseline, the percentage with a usual SBP >or=140 mm Hg was 31.6% in the UK-TIA trial, 48.2% in the Dutch TIA trial, and 57.7% in the ECST. At least 3 consecutive measurements of SBP <120 mm Hg were required to be >90% certain that subsequent usual SBP would not be >or=140 mm Hg. CONCLUSIONS: Given the greater medium-term variability of BP in patients with a previous TIA or stroke than in the general population, single measurements of "normal" or "low" BP will substantially underestimate the true prevalence of hypertension.