Incorporating Prior Data in Quantitative Benefit-Risk Assessments: Case Study of a Bayesian Method.

BACKGROUND: Multiple criteria decision analysis (MCDA) and stochastic multi-criteria acceptability analysis (SMAA) in their current implementation cannot incorporate prior or external information on benefits and risks. We demonstrate how to incorporate prior data using a Bayesian mixture model approach while conducting quantitative benefit-risk assessments (qBRA) for medical products. METHODS: We implemented MCDA and SMAA in a Bayesian framework. To incorporate information from a prior study, we use mixture priors on each benefit and risk attribute that mixes information from a previous study with a vague prior distribution. The degree of borrowing is varied using a mixing proportion parameter. RESULTS: A demonstration case study for qBRA using the supplementary New Drug Application (sNDA) filing for Rivaroxaban for the indication of reduction in the risk of major thrombotic vascular events in patients with peripheral artery disease (PAD) was used to illustrate the method. Net utility scores, obtained from the randomized controlled trial data to support the sNDA, from the MCDA for Rivaraxoban and comparator were 0.48 and 0.56, respectively, with Rivaroxaban being the preferred alternative only 33% of the time. We show that with only 30% borrowing from a previous RCT, the MCDA and SMAA results are favorable for Rivaroxaban, accounting for the seemingly aberrant results on all-cause death in the trial data used to support the sNDA. CONCLUSION: Our method to formally incorporate prior data in MCDA and SMAA is easy to use and interpret. Software in the form of an RShiny App is available here: https://sai-dharmarajan.shinyapps.io/BayesianMCDA_SMAA/ .

Consideration of the adaptive randomization allocation ratio in the presence of treatment group heteroscedasticity in clinical trials.

For randomized clinical trials, subjects' variance structures may vary over time among treatment groups, resulting in the heteroscedasticity of residuals in a regression analysis. Commonly used methods that assume equal variance among all treatment groups may not be able to control for a type I error. When the variances are indeed the same across treatment groups, an equal randomization allocation ratio will yield the greatest study power. However, out of ethical concern or urgent need for rare disease clinical trials, more patients may have to be allocated to the study drug arm. In these situations, an unequal randomization ratio should be considered. We propose a group variance-covariance and structures-based method to adapt the randomization ratio after interim analysis. We use simulations to compare commonly used statistical methods for continuous endpoints in assessing the impact of heteroscedasticity in equal and unequal randomization ratios and examine the extent to which the findings are affected by missing data.

Challenges and potential strategies utilizing external data for efficacy evaluation in small-sized clinical trials.

In clinical trials for diseases with very small patient populations, trial investigators may encounter recruitment difficulties. It can be challenging to conduct clinical trials with enough power to detect a treatment effect, and randomization may not be feasible due to timeline, budget, and ethical concerns. To bring breakthrough therapies to the market quickly, it is important to come up with efficient approaches to utilizing individual patient data through improved study design and sound statistical methods. Emerging topics in this area include the use of Bayesian approaches to flexibly incorporate prior information into the current clinical trials, the use of historical controls to efficiently conduct trials that will reduce the number of subjects recruited and ease ethical considerations, and the use of innovative study designs, such as a platform design, to improve the efficiency and speed of the medical therapy development progress. In this paper, we describe three scenarios which highlight some of the challenges encountered in small-sized clinical trial development and provide potential statistical approaches to overcome the aforementioned challenges.

Utilizing shared internal control arms and historical information in small-sized platform clinical trials.

Recruitment of patients in concurrent control arms can be very challenging for clinical trials for pediatric and rare diseases. Innovative approaches, such as platform trial designs, including shared internal control arm(s), can potentially reduce the needed sample size, improving the efficiency and speed of the drug development program. Furthermore, historical borrowing, which involves leveraging information from control arms in previous relevant clinical trials, may further enhance a clinical trial's efficiency. In this paper, we discuss platform trials highlighting their advantages and limitations. We then compare various strategies that borrow historical data or information, such as pooling data from different studies, analyzing data from studies separately, test-then-pool, dynamic pooling, and Bayesian hierarchical modeling, which focuses on the meta-analytic-predictive (MAP) prior. We further propose a procedure to illustrate the feasibility of utilizing historical controls under a platform setting and describe the statistical performance of our method via simulations.

The plan of enrichment designs for dealing with high placebo response.

To deal with high placebo response in clinical trials for psychiatric and other diseases, different enrichment designs, such as the sequential parallel design, two-way enriched design, and sequential enriched design, have been proposed and implemented recently. Depending on the historical trial information and the trial sponsors' resources, detailed design elements are needed for determining which design to adopt. To assist in making more suitable decisions, we perform evaluations for selecting required design elements in terms of power optimization and sample size planning. We also discuss the implementation of the interim analysis related to its applicability.

Unified additional requirement in consideration of regional approval for multiregional clinical trials.

To speed up the process of bringing a new drug to the market, more and more clinical trials are being conducted simultaneously in multiple regions. After demonstrating the overall drug's efficacy across regions, the regulatory and drug sponsor may also want to assess the drug's effect in specific region(s). Most of the recent approaches imposed a uniform criterion to assess the consistency of treatment effects between the interested region(s) and the entire study population regardless of the number of regions in multiregional clinical trials (MRCT). As a result, the needed sample size to achieve the desired probability of satisfying the regional requirement could be huge and implausible for the trial sponsors to implement. In this paper, we propose a unified additional requirement for regional approval by differing the parameters in the additional requirement depending on the number of planned regions. In particular, the values of the parameters are determined by a reasonable sample size increase with the desired probability satisfying the additional requirement. Considering the practicality of the global trial or sample size increase, we recommend specific values of the parameters for a different number of planned regions. We also introduce the assurance probability curve to evaluate the performance of different regional requirements.

Clinical endpoints and adaptive clinical trials in precirrhotic nonalcoholic steatohepatitis: Facilitating development approaches for an emerging epidemic.

Due to the increasing prevalence of nonalcoholic steatohepatitis (NASH) and its associated health burden, there is a high need to develop therapeutic strategies for patients with this disease. Unfortunately, its long and asymptomatic natural history, the uncertainties about disease progression, the fact that most patients are undiagnosed, and the requirement for sequential liver biopsies create substantial challenges for clinical development. Adaptive design methods are increasingly used in clinical research as they provide the flexibility and efficiency for identifying potential signals of clinical benefit of the test treatment under investigation and make prompt preplanned adaptations without undermining the validity or integrity of the trial. Given the high unmet medical need and the lack of validated surrogate endpoints in NASH, the use of adaptive design methods appears reasonable. Furthermore, due to the limited number of patients willing to have multiple liver biopsies and the need for long-term exposure to assess an impact in outcomes, a continuous seamless adaptive design may reduce the overall sample size while allowing patients to continue after each one of the phases. Here, we review strategic frameworks that include potential surrogate endpoints as well as statistical and logistical approaches that could be considered for applying adaptive designs to clinical trials in NASH with the goal of facilitating drug development for this growing medical need. (Hepatology Communications 2017;1:577-585).

Testing treatment effect in schizophrenia clinical trials with heavy patient dropout using latent class growth mixture models.

By examining the outcome trajectories of the dropout patients with different reasons in the schizophrenia trials, we note that although patients are recruited from the same protocol that have compatible baseline characteristics, they may respond differently even to the same treatment. Some patients show consistent improvement while others only have temporary relief. This creates different patient subpopulations characterized by their response and dropout patterns. At the same time, those who continue to improve seem to be more likely to complete the study while those who only experience temporary relief have a higher chance to drop out. Such phenomenon appears to be quite general in schizophrenia clinical trials. This simultaneous inhomogeneity both in patient response as well as dropout patterns creates a scenario of missing not at random and therefore results in biases when we use the statistical methods based on the missing at random assumption to test treatment efficacy. In this paper, we propose to use the latent class growth mixture model, which is a special case of the latent mixture model, to conduct the statistical analyses in such situation. This model allows us to take the inhomogeneity among subpopulations into consideration to make more accurate inferences on the treatment effect at any visit time. Comparing with the conventional statistical methods such as mixed-effects model for repeated measures, we demonstrate through simulations that the proposed latent mixture model approach gives better control on the Type I error rate in testing treatment effect. Published 2016. This article is a U.S. Government work and is in the public domain in the USA.

A small n sequential multiple assignment randomized trial design for use in rare disease research.

BACKGROUND: Clinical trials in rare diseases are difficult to conduct due to the limited number of patients available with each disorder. We developed a Phase 2 trial which is a small n sequential multiple assignment randomized trial (snSMART) design to test several treatments for a rare disease for which no standard therapy exists. PURPOSE: This paper illustrates the design, sample size estimation and operating characteristics of an snSMART. METHODS: We investigate the performance of a class of weighted Z statistics via computer simulations. RESULTS: We demonstrate the increase in power over traditional single stage designs, and indicate how the power changes as a function of the weight given to each stage. CONCLUSION: The snSMART design is promising in a rare disease setting where several alternative treatments are under consideration and small sample sizes are necessary.

Review of maintenance trials for major depressive disorder: a 25-year perspective from the US Food and Drug Administration.

OBJECTIVE: The maintenance efficacy of antidepressants is usually assessed in postmarketing studies with a randomized withdrawal design. This report explores differences in relapse rates, trial characteristics, and success rates in maintenance efficacy studies submitted to the US Food and Drug Administration (FDA) over a 25-year period. DATA SOURCES: Clinical data from all maintenance trials with antidepressants submitted to FDA between 1987 and 2012. STUDY SELECTION: Efficacy data were compiled from 15 maintenance clinical trials in adults diagnosed with major depressive disorder according to DSM-III or DSM-IV criteria. DATA EXTRACTION: Trial characteristics, relapse rates, and time to relapse in each study were examined. RESULTS: Relapse rates were significantly lower (P < .05) in the drug arm than in the placebo arm in every study, with a mean relapse rate difference of 18% and an average percent reduction in relapse rate of 52% compared to placebo. Only 6% of the relapse events occurred in the first 2 weeks of the double-blind phase. The separation between treatment arms continued to increase throughout the double-blind phase only in the trial with longest response stabilization period. CONCLUSIONS: Antidepressant maintenance trials have a high rate of success, indicating a benefit of continuing drug treatment after initial response to an antidepressant. This benefit appears to result mainly from a decreased rate of recurrent depression rather than from an effect of drug withdrawal in the placebo groups.

A sequential enriched design for target patient population in psychiatric clinical trials.

High placebo response is widely believed to be one major reason why many psychiatric clinical trials fail to demonstrate drug efficacy. In order to alleviate this problem, research has developed several enrichment designs, including the parallel design with a placebo lead-in phase, the sequential parallel design, and a recently proposed two-way enriched design. While these designs have been evaluated and discussed individually, their effectiveness against each other has not been rigorously compared. The current study examines the various enrichment designs simultaneously. Building on their strengths, we introduce a new improved design named' sequential enriched design' (SED) aimed at removing not only patients with high placebo response but also patients who do not respond to any treatment from the study. The SED begins with a double-blind placebo lead-in phase followed by a traditional parallel design in the first stage. Only patients who respond to the drug in the first stage are re-randomized to the drug or placebo at the second stage. We simulate data for a mixed population composed of four subgroups of patients who are predetermined as to whether they respond to drug or not as well as to placebo or not. By focusing on the target patients whose responses reflect the drug's efficacy,we evaluate the bias, mean squared error, and power for different designs. We demonstrate that the SED produces a less biased estimate for the target treatment effect and yields reasonably high power in general compared with the other designs.

Exploratory analyses of efficacy data from schizophrenia trials in support of new drug applications submitted to the US Food and Drug Administration.

OBJECTIVE: There has been concern about a high rate of placebo response and a decline in treatment effect over time in schizophrenia trials as well as the implications of increasing conduct of such trials outside North America. This report explores differences in efficacy data over an 18-year period from randomized placebo-controlled trials submitted in support of new drug applications (NDAs) for the treatment of schizophrenia and differences in results between trials conducted in North America and elsewhere. DATA SOURCES: Clinical trial data that were submitted to the US Food and Drug Administration (FDA) as part of NDAs for the indication of schizophrenia between 1991 and 2009. STUDY SELECTION: Efficacy data were compiled from 32 clinical trials with 11,567 evaluable patients with schizophrenia. Data from completed, randomized, multicenter, double-blind, placebo-controlled, 4- to 8-week clinical trials in adult patients diagnosed with schizophrenia according to DSM-III or DSM-IV criteria were included. DATA EXTRACTION: Baseline demographic and disease characteristics, including mean Positive and Negative Syndrome Scale (PANSS) total scores, were summarized and compared between North American and multiregional trials. Mean change from baseline to endpoint in PANSS total scores was utilized as the primary outcome of interest. We explored differences in treatment effect and success rate of these trials based on when and where the studies were conducted, sample size, trial duration, and baseline patient characteristics. RESULTS: Twenty-one of the 32 trials were conducted solely in North America, and 11 were carried out in multiple regions. Of those 11 multiregional trials, 2 were conducted exclusively in foreign countries. Although the observed responses (change from baseline) in placebo and drug-treated groups in multiregional trials tended to be larger than in North American trials, the treatment effects (drug-placebo difference) were -9 and -8 PANSS units for North American and multiregional trials, respectively. When time of trial conduct was taken into account, an increasing placebo response and a diminishing treatment effect over time were observed in North American trials from -10.8 PANSS units for the first period (1991-1998) to -6.0 PANSS units for the later period (1999-2008). The overall trial success rate over the almost 2 decades was 78%, declining slightly in trials conducted after 1999, the time period during which multiregional trials were first conducted (74% for 1999-2008 vs 85% for 1991-1998), despite increasing sample sizes in the later period. The mean baseline PANSS total score was in the range of 87-100 for most of these trials. Trials in patients with higher mean baseline PANSS total scores tended to show larger treatment effects than those in patients with lower scores. The mean body weight and body mass index (BMI) were higher in patients in North American trials and North America-predominant multiregional trials compared to those in foreign-predominant multiregional trials (mean body weights of 85 kg and 81 kg vs 72 kg, and BMIs of 29 and 27 vs 25, respectively). Treatment effects decreased as body weights increased, especially in North American trials. In foreign-predominant multiregional trials, there were higher proportions of women than in North American trials and North America-predominant multiregional trials (40% vs 22% and 27%, respectively) and a relatively larger proportion of Asians (21% vs 1% and 8%, respectively). CONCLUSIONS: A high and increasing placebo response and a declining treatment effect are of great concern in schizophrenia trials conducted in North America. In this era of global clinical trials, close attention is needed to the design and conduct of these trials.

Evaluation of performance of some enrichment designs dealing with high placebo response in psychiatric clinical trials.

Dealing with high placebo response remains a big challenge to conventional clinical trials for psychiatric disorders. A widely-used design strategy is to implement a placebo lead-in phase prior to randomization. The sequentially parallel design (SPD) proposed by Fava et al., which contains two consecutive double-blind treatment stages, has recently been promoted to reduce both the high placebo response and the required sample size in clinical trials for psychiatric disorders. Our work aims to study these two design strategies and evaluate the relevant statistical approaches for continuous measures under SPD in the presence of missing data. Based on the FDA archived database, we found that a longer placebo lead-in period seemed to help in identifying more placebo responders and thus increase the chance to detect a drug-placebo difference on continuous efficacy endpoint. Using a simple weighted ordinary least square test statistic Z(OLS), we analytically showed that, under the SPD with re-randomization of placebo non-responders at the second stage (SPD-ReR), Z(OLS) can be used as a viable alternative to the weighted test statistic based on seemingly unrelated regression estimate Z(SUR) proposed by Tamura and Huang to assess treatment efficacy. Results from simulation study comparing three imputation methods (last-observation-carried-forward approach, multiple imputation, and mixed-effects model for repeated measures (MMRM)) demonstrate that, when data are missing-at-random under SPD-ReR and the dropout rate is moderate, the weighted test statistic based on MMRM estimates appears to be the most robust test statistic for SPD-ReR in terms of type I error control, power performance, and estimation accuracy.

Exploratory analyses of efficacy data from major depressive disorder trials submitted to the US Food and Drug Administration in support of new drug applications.

OBJECTIVE: There has been concern about a high rate of placebo response and a substantial failure rate in recent clinical trials in major depressive disorder (MDD). This report explores differences in efficacy data from placebo-controlled MDD trials submitted in support of new drug applications (NDAs) over a 25-year period. METHOD: We compiled efficacy data from 81 randomized, double-blind clinical trials, with 21,611 evaluable patients, that were submitted to the US Food and Drug Administration as part of NDAs for an antidepressant claim between 1983 and 2008. Trial data were limited to completed, randomized, multicenter, double-blind, placebo-controlled clinical trials in adult patients diagnosed with MDD according to DSM-III or DSM-IV criteria. The database was further limited to patients who were involved in clinical trials for drugs widely viewed as effective antidepressants and for doses of these drugs also viewed as effective doses. Trials were rated as successful if they showed statistical superiority vs placebo for the investigational drug on change in Hamilton Depression Rating Scale (HDRS) score (last-observation-carried-forward data). (Trials with multiple investigational drug groups were successful if there was superiority in at least 1 drug group after adjustment for multiplicity.) In particular, we explored differences in effect size and success rate of these trials, based on when the studies were conducted, geographic location of the study sites (US vs non-US), trial duration, dosing regimen, study size, and baseline disease characteristics. RESULTS: Eighty-one percent of MDD patients were enrolled in US sites. Although the observed placebo and drug responses at non-US sites tended to be larger than at US sites, the treatment effect (drug-placebo difference) was similar (mean change from baseline of about -2.5 units in HDRS total score) in US and non-US trials. In both US and non-US trials, the placebo response showed a modest increase over the observation period (1983-2008). Treatment effect clearly diminished over this same period, at a similar rate for both US and non-US trials despite a marked increase in the sample size of the trials. Our analysis showed that 53% of all MDD trials in the last 25 years were successful. US trials had a higher success rate than non-US trials (58% vs 33%). Before 1995, the overall success rate was 55%, compared to 50% for trials in 1995 or later, and, in general, 6-week trials had a higher success rate than 8-week trials (55% vs 42%). It should be noted that the earlier trials were mostly 6 weeks, and the 6-week trials had higher mean baseline HDRS scores than the 8-week trials. Study size did not seem to influence trial success rates. Mean baseline HDRS total scores declined over the 25-year observation period for patients in both US and non-US trials, as did treatment effect in these trials, again, regardless of region. Fixed-dose trials had a numerically slightly greater success rate than flexible-dose trials (57% vs 51%), although on average treatment effect was numerically larger in the flexible-dose trials than in fixed-dose trials (mean of -2.9 vs -2.0 on HDRS units). CONCLUSIONS: Treatment effect has declined over time in MDD trials, and there has been a high failure rate for these trials during the entire period, but the reasons for these findings remain elusive. Baseline disease severity seems to be a more important factor in study outcome than study duration, dosing regimen, sample size, time when studies were conducted, and regions where data were generated. Close attention is needed to a variety of factors in the design and conduct of these studies, including patient population, diagnostic considerations, patient assessment, and clinical practice differences. These considerations become increasingly important as globalization of clinical trials continues to increase.

Trial design issues and treatment effect modeling in multi-regional schizophrenia trials.

In recent years, we have seen an increasing trend of foreign data as part of clinical trial data submitted in new drug applications (NDA) to US Food and Drug Administration (FDA). To understand the design and analysis characteristics, we studied schizophrenia multi-regional clinical trials (MRCTs). The schizophrenia data set consisted of a total of 12,585 patients collected from 33 clinical trials with 63.8% patients from North America, the largest region. The data set constituted 10 schizophrenia drug programs in support of NDAs submitted to FDA from December 1993 to December 2005. Two main objectives were pursued. First, we investigated some study design issues including potential heterogeneity of treatment effect via meta analysis and placebo response pattern over time. Second, we performed empirical modeling in two ways, supervised and unsupervised, to explain potential impact of baseline covariates on treatment effect in MRCTs. Based on our analysis results, placebo response appeared to increase over time and primarily attributed to US region. On average, the observed treatment effect in the US was generally smaller than non-US region. Both supervised and unsupervised empirical modeling selected baseline Positive and Negative Syndrome Scale total score as one of the most important covariates explaining a treatment effect. Region also played a role in explaining potential treatment effect heterogeneity. When baseline body weight was considered as a covariate in an empiric model, our results indicated that it alone did not seem to be an important factor in explaining regional difference.

A bias correction in testing treatment efficacy under informative dropout in clinical trials.

In clinical trials of drug development, patients are often followed for a certain period of time, and the outcome variables are measured at scheduled time intervals. The main interest of the trial is the treatment efficacy at a prespecified time point, which is often the last visit. In such trials, patient dropout is often the major source for missing data. With possible informative patient dropout, the missing information often causes biases in the inference of treatment efficacy. In this article, for a time-saturated treatment effect model and an informative dropout scheme that depends on the unobserved outcomes only through the random coefficients, we propose a grouping method to correct the biases in the estimation of treatment effect. The asymptotic variance estimator is also obtained for statistical inference. In a simulation study, we compare the new method with the traditional methods of the observed case (OC) analysis, the last observation carried forward (LOCF) analysis, and the mixed model repeated measurement (MMRM) approach, and find it improves the current methods and gives more stable results in the treatment efficacy inferences.

Approval summary for erlotinib for treatment of patients with locally advanced or metastatic non-small cell lung cancer after failure of at least one prior chemotherapy regimen.

PURPOSE: To describe the Food and Drug Administration (FDA) review and approval of erlotinib (Tarceva, OSI Pharmaceuticals, Melville, NY) for treatment of patients with locally advanced or metastatic non-small cell lung cancer (NSCLC) after failure of at least one prior chemotherapy regimen. EXPERIMENTAL DESIGN: The FDA reviewed raw data in electronic format from a randomized controlled clinical trial comparing erlotinib with placebo in patients with locally advanced or metastatic NSCLC after failure of at least one prior chemotherapy regimen. RESULTS: Patients were randomized in a 2:1 ratio (erlotinib, n = 488 and placebo, n = 243). Erlotinib was superior to placebo for survival, progression-free survival, and tumor response rate. Exploratory analyses indicate that epidermal growth factor receptor status may be an important predictor of the erlotinib survival effect. Rash (75% versus 17%) and diarrhea (54% versus 18%) in the erlotnib and placebo group respectively were the most common adverse events. Severe rash occurred in 9% and severe diarrhea in 6% of erlotinib-treated patients and each resulted in study discontinuation in 1% of patients. Dose reductions were required for 10% of patients with rash and 4% of patients with diarrhea. CONCLUSIONS: On November 18, 2004, the FDA granted erlotinib regular approval for treatment of patients with locally advanced or metastatic NSCLC after failure of at least one prior chemotherapy regimen. The applicant has committed to conduct post-marketing clinical trials to assess further the effect of epidermal growth factor receptor expression, measured with immunohistochemical staining, on erlotinib treatment effect.

FDA drug approval summary: erlotinib (Tarceva) tablets.

On November 18, 2004, erlotinib (Tarceva); OSI Pharmaceuticals, Inc., Melville, NY, http://www.osip.com, and Genentech, Inc., South San Francisco, CA, http://www.gene.com) received regular approval as monotherapy for the treatment of patients with locally advanced or metastatic non-small cell lung cancer (NSCLC) after failure of at least one prior chemotherapy regimen. Survival of erlotinib-treated patients was superior to that of placebo-treated patients. The median survival duration of erlotinib-treated patients was 6.67 months, compared with 4.70 months for placebo-treated patients. Exploratory univariate analyses showed a larger survival prolongation in two subsets of patients: those who never smoked and those with epidermal growth factor receptor (EGFR)-positive tumors. Patients who never smoked and were EGFR-positive had a large erlotinib survival benefit. Erlotinib was also superior to placebo for progression-free survival and a response rate of 8.9% versus 0.9%. Skin rash and diarrhea were the most common erlotinib adverse events. Severe rash occurred in 8%, and severe diarrhea occurred in 6% of erlotinib-treated patients. In the first-line treatment of NSCLC, two large, controlled, randomized trials showed no benefit from adding erlotinib to doublet, platinum-based chemotherapy. Therefore, erlotinib is not indicated for use in this setting.