In clinical trials, almost all key milestone dates can be defined in terms of time to endpoint maturation (TTEM). The real time monitoring and accurate prediction of TTEM have a significant impact on clinical trial planning and execution and can bring significant value to clinical trial practitioners. TTEM is defined as the time to achieve or observe a certain number or percentage of some endpoint of interest. It is a combination of time to site initiation, time to subject enrollment after site initiation and time to event of interest after subject enrollment. To better predict TTEM during the trial, the future site initiation and subject enrollment have to be taken into account while predicting the number of events. In this article, we propose a novel simulation-based framework combining time to site initiation, time to subject enrollment and time to event in order to predict TTEM. A nonhomogeneous Poisson process with a quadratic time-varying rate function is used to model site initiation and subject enrollment and more advanced time to event models had been explored and integrated on top of them, such as Weibull, piecewise exponential, and model averaging which is equivalent to a Bayesian model selection strategy. To evaluate the predictive performance of the proposed methodology, we conducted extensive simulations and applied the methodology to 14 randomly selected real oncology phase 2 and phase 3 studies in both solid tumor and hematology with a total 31 study-endpoint combinations. The predictive performance of the proposed methodology was then compared with popular and commonly available commercial software, for example, East (Cytel, Cambridge, MA, USA). From both simulation and real data, the proposed methodology can significantly improve the prediction accuracy by up to 54% compared to the commonly available method.

Decision-making is central to every phase of drug development, and especially at the proof of concept stage where risk and evidence must be weighed carefully, often in the presence of significant uncertainty. The decision to proceed or not to large expensive Phase 3 trials has significant implications to both patients and sponsors alike. Recent experience has shown that Phase 3 failure rates remain high. We present a flexible Bayesian quantitative decision-making paradigm that evaluates evidence relative to achieving a multilevel target product profile. A framework for operating characteristics is provided that allows the drug developer to design a proof-of-concept trial in light of its ability to support decision-making rather than merely achieve statistical significance. Operating characteristics are shown to be superior to traditional p-value-based methods. In addition, discussion related to sample size considerations, application to interim futility analysis and incorporation of prior historical information is evaluated.

BACKGROUND & AIMS:MEDI2070 is a human monoclonal antibody that selectively inhibits interleukin 23 (IL23), a cytokine implicated in the pathogenesis of Crohn's disease (CD). We analyzed its safety and efficacy in treatment of CD in a phase 2a study.METHODS:We conducted a double-blind, placebo-controlled study of 119 adults with moderate to severe CD failed by treatment with tumor necrosis factor antagonists. Patients were randomly assigned (1:1) to groups given MEDI2070 (700 mg) or placebo intravenously at weeks 0 and 4. Patients received open-label MEDI2070 (210 mg) subcutaneously every 4 weeks from weeks 12 to 112. The CD Activity Index was used to measure disease activity.RESULTS:The primary outcome, clinical response (either a 100-point decrease in CD Activity Index score from baseline or clinical remission, defined as CD Activity Index score <150) at week 8 occurred in 49.2% of patients receiving MEDI2070 (n = 59) compared with 26.7% receiving placebo (n = 60; absolute difference, 22.5%; 95% confidence interval, 5.6%-39.5%; P = .010). Clinical response at week 24 occurred in 53.8% of patients who continued to receive open-label MEDI2070 and in 57.7% of patients who had received placebo during the double-blind period and open-label MEDI2070 thereafter. The most common adverse events were headache and nasopharyngitis. Higher baseline serum concentrations of IL22, a cytokine whose expression is induced by IL23, were associated with greater likelihood of response to MEDI2070 compared with placebo.CONCLUSIONS:In a phase 2a trial of patients with moderate to severe Crohn's disease who had failed treatment with tumor necrosis factor antagonists, 8 and 24 weeks of treatment with MEDI2070 were associated with clinical improvement. ClinicalTrials.gov ID: NCT01714726.

Subgroup identification for personalized medicine has become very popular in the last decade. Efficient recursive partitioning procedures adapted from machine learning are natural approaches for performing subgroup identification based on pre-defined biomarkers since they provide subgroups as terminal nodes in the decision tree. However, recursive partitioning is also known as a potentially unstable procedure with results being quite sensitive to normal sampling variability in the data. One common approach, borrowed from ensemble learning, to overcome such instability is application of recursive partitioning to multiple data sets sampled from the observed data followed by averaging the results over the collection of subgroups. This article proposes an alternative approach to subgroup identification in clinical trials that first evaluates the predictive strength of biomarkers based on variable importance and then applies recursive partitioning to the biomarkers with the highest variable importance scores. A deterministic version of this idea was implemented in the Adaptive SIDEScreen method that generates a collection of patient subgroups by retaining multiple candidate splits of each parent group by different biomarkers (Lipkovich and Dmitrienko 2014a, 2014b). Now, we extend the Adaptive SIDEScreen and introduce the Stochastic SIDEScreen method. The key idea is to introduce randomness in the subgroup generation process, borrowing from bagging methods, to produce a broader collection of subgroups. Specifically, the SIDES method, where the most promising biomarkers are selected for each parent group from a set of candidate biomarkers, is applied to multiple bootstrap samples of the data. This new approach leads to a more reliable biomarker selection process, which is especially important for smaller, early phase studies when biomarker selection is typically carried out. The method is illustrated using clinical trial examples.

BACKGROUND:Neuromyelitis optica spectrum disorder (NMOSD) is a rare, disabling autoimmune disorder of the central nervous system. Clinical trials in NMOSD present unique design and statistical challenges to adequately determine treatment effect and to minimize risk.METHODS:The N-MOmentum trial (NCT02200770) is evaluating the efficacy and safety of MEDI-551, an anti-CD 19 B-cell depleting monoclonal antibody, in patients with NMOSD and employs a number of unique design features. Patients are randomized (3:1) to receive MEDI-551 or placebo for up to 197 days. NMOSD attacks are evaluated by the investigator and confirmed by an independent adjudication committee. The primary endpoint is time to first relapse as determined by adjudication committee. Sample size re-estimation and futility analyses are planned interim analyses. Novel multiplicity adjustment methods are developed to control the study-wise type I error. Methods for assessing inter- and intrarater reliability are proposed.CONCLUSIONS:The N-MOmentum study minimizes exposure to placebo for individual patients. The application of several statistical methods in the N-MOmentum trial is novel in NMOSD and aims to achieve a balance between minimizing risk and maintaining scientific integrity.

In study designs for randomized clinical trials with a survival endpoint, the log-rank test is commonly used with the treatment effect under a proportional hazards assumption. Recently, treatment effects in cancer immunotherapy trials have exhibited a delayed effect pattern with late separation of survival curves, raising challenges to the use of conventional study design hypotheses and analysis. In particular, when a trial with interim analyses is designed using a group sequential method, the expected treatment effect from a log-rank test statistic varies across analysis times and differs from the parameter specified under the alternative hypothesis. In this article, we present statistical analytical work that formulates a design including interim analyses with a survival endpoint under a delayed treatment effect alternative. Closed-form solutions are provided for calculating power and sample size over varying study/follow-up times for the group sequential, delayed treatment effect design. The analytical work is also presented graphically and simulations are conducted for validation.

Adaptive design clinical trial methodologies offer both opportunities and challenges for observing basic ethical principles in human subject research. Using both published and unpublished adaptive design clinical trials, we have selected and reviewed examples of clinical trials with different design adaptations to discuss the ethical obstacles presented and often successfully resolved by these approaches, including (1) confirmatory trials for treatments widely accepted on the basis of uncontrolled case series or open-label trials (clinical equipoise and “justice” in the sense of which trial groups will “receive the benefits of research and bear its burdens”) (infantile hemangioma/propranolol); (2) interim results analysis by unblinded data monitoring committees (“withholding information necessary to make a considered judgment” [“respect for persons”] versus compromising the trial’s scientific basis) (BIG 1-98); (3) adaptations involving sample size reassessment or dose adjustment via dropping or adding treatment arms, allowing fewer subjects to produce statistically significant results, fewer subjects treated with ineffective/toxic doses, and more subjects given doses showing tolerance and treatment activity (“beneficence” or “protecting from harm and making efforts to secure wellbeing”) (ECMO, Neuromyelitis Optica); (4) adaptive randomization inferential problems balanced against ethical benefits (trastuzumab vs taxane in advanced gastric cancer; ADVENT); (5) more efficient allocation of societal resources for research, in both public and commercial realms, versus uncertain regulatory acceptance (indicaterol; VALOR); and (6) platform, umbrella, and basket trials offering additional efficiencies (I-SPY II, BATTLE, Lung-MAP). The importance of careful design, meticulous planning, and rigorous ethical review of adaptive design trials on a case-by-case basis cannot be overemphasized.

Objectives Genotype-specific associations between hepatitis C virus (HCV) and insulin resistance (IR) have been described, but a causal relationship remains unclear. This study investigated the association between a sustained virological response (SVR) and IR after chronic HCV therapy.Methods 2255 treatment-naive patients with chronic HCV genotype 1 or 2/3 were enrolled in two phase 3 trials of albinterferon alpha-2b versus pegylated interferon alpha-2a for 48 or 24 weeks, respectively. IR was measured before treatment and 12 weeks after treatment using homeostasis model assessment (HOMA)-IR.Results Paired HOMA-IR measurements were available in 1038 non-diabetic patients (497 with genotype 1; 541 with genotype 2/3). At baseline the prevalence of HOMA-IR >3 was greater in patients with genotype 1 than 2/3 (33% vs 27%; p=0.048). There was a significant reduction in the prevalence of IR in patients with genotype 1 achieving SVR (delta 10%; p<0.001), but not in genotype 1 non-responders or those with genotype 2/3. Multivariate analysis indicated that SVR was associated with a significant reduction in mean HOMA-IR in patients with genotype 1 (p=0.004), but not in those with genotype 2/3, which was independent of body mass index, alanine transaminase, gamma-glutamyl transpeptidase and lipid level changes.Conclusions SVR is associated with a reduction in HOMA-IR in patients with HCV genotype 1 but not in those with genotype 2/3. Genotype 1 may have a direct effect on the development of IR, independent of host metabolic factors, and may be partially reversed by viral eradication.

CONTEXT:Thyroid dysfunction is a common complication of interferon-α (IFNα) therapy, with many phenotypic patterns and the potential for significant morbidity. OBJECTIVE:Our objective was to gain mechanistic insight and predict clinical presentations by determining the risk factors for distinct subtypes of IFNα-induced thyroid dysfunction. DESIGN:ACHIEVE-1, a randomized trial conducted from 2005-2009, compared long-acting preparations of IFNα in 1323 patients with hepatitis C, genotype 1. SETTING:A total of 149 outpatient clinics in North America, Europe, and Australia participated. PATIENTS:We studied 1233 patients who were euthyroid at baseline. This population is 60% male and 82% Caucasian. INTERVENTIONS:Patients were treated with pegylated IFNα2a weekly or albumin-IFNα2b every 2 wk for 48 wk. Serum TSH and free T(4) were measured before therapy and 12 or more times over 60 weeks. MAIN OUTCOME MEASURES:Thyroid dysfunction was defined as a TSH outside the normal range during the course of therapy. Low serum TSH indicated thyrotoxicosis, elevated TSH indicated hypothyroidism, and both abnormalities occurred in biphasic thyroiditis. RESULTS:Of previously euthyroid patients, 16.7% developed abnormal TSH values during therapy, including 24 with TSH below 0.1 mU/liter, 69 with TSH over 5.5 mU/liter, and 76 with biphasic thyroiditis. Biphasic thyroiditis was over 8-fold more common among women than men using multivariate logistic regression analysis [odds ratio (OR) = 8.4; 95% confidence interval (CI) = 4.5-15.8]. Thyrotoxicosis was most strongly associated with a lower pretreatment TSH (OR = 4.1 per -1 mU/liter decline; 95% CI = 1.9-9), whereas hypothyroidism was strongly associated with higher pretreatment TSH (OR = 3.9 per 1 mU/liter increase; 95% CI = 3-5.2). CONCLUSIONS:Biphasic thyroiditis is common among women treated for hepatitis C with IFNα. Lower and higher pretreatment serum TSH are associated with greater likelihood of thyrotoxicosis and hypothyroidism, respectively. Antithyroid antibody levels were not available for the cohort, and thus we cannot clarify the role of pretreatment thyroid autoimmunity as a risk factor. Our results do show that readily identifiable patient characteristics are risk factors for specific patterns of IFN-induced thyroid dysfunction. These findings suggest that distinct mechanisms may underlie subtypes of thyroid dysfunction associated with immune-modulatory therapy for hepatitis C.

AIM:To compare histological endpoint assessment using noninvasive alternatives to biopsy during treatment in a chronic hepatitis C virus (HCV) cohort.METHODS:Patients with chronic HCV were randomized to receive interferon-based therapy for 24 (genotypes 2/3) or 48 (genotype 1) wk. FibroSURE™ (FS) was assessed at baseline and at week-12 post-treatment follow-up. Baseline biopsy for METAVIR was assessed by a single pathologist. FibroScan(®) transient elastography (TE) was performed during treatment in a patient subset.RESULTS:Two thousand and sixty patients (n = 253 in Asia) were classified as METAVIR F0-1 (n = 1682) or F2-4 (n = 378). For F2-4, FS (n = 2055) had sensitivity and specificity of 0.87 and 0.61, respectively, with area under the receiver-operating curve of 0.82; corresponding values for TE (n = 214) and combined FS/TE (n = 209) were 0.77, 0.88 and 0.88, and 0.93, 0.68 and 0.88. Overall FS/TE agreement for F2-4 was 71% (κ = 0.41) and higher in Asians vs non-Asians (κ = 0.86 vs 0.35; P < 0.001). Combined FS/TE had 97% accuracy in Asians (n = 33). Baseline FS (0.38 vs 0.51, P < 0.001) and TE (8.0 kPa vs 11.9 kPa, P = 0.006) scores were lower in patients with sustained virological response than in nonresponders, and were maintained through follow-up.CONCLUSION:FS and TE may reliably differentiate mild from moderate-advanced disease, with a potential for high diagnostic accuracy in Asians with chronic HCV.

Background and Aim: The role of insulin resistance (IR) and hepatic steatosis in fibrogenesis in chronic hepatitis C infection (CHC) has yielded conflicting data and few studies have been performed in Asian-region populations. We retrospectively investigated the relationship between host metabolic variables, including IR and hepatic steatosis, to hepatic fibrosis in Asian-region CHC genotype 2/3 patients.Methods: A total of 303 treatment-naive Asian-region patients with CHC genotype 2/3 were enrolled in a multicenter phase 3 study of albinterferon alfa-2b plus ribavirin for 24 weeks. IR was defined as Homeostasis Model for Assessment of IR (HOMA-IR) > 2. Baseline liver biopsy was evaluated by a single expert histopathologist. Post hoc subgroup logistic regression modeling selected for independent variables associated with significant fibrosis (METAVIR stage F2-F4).Results: Insulin resistance was available in 263 non-diabetic Asian-region patients (hepatitis C virus-2 [HCV-2] = 171, HCV-3 = 92), and 433 non-Asian region patients (407 "Caucasian"); METAVIR fibrosis prevalence F0-F1 (minimal fibrosis) = 201 (77%) and F2-F4 (significant fibrosis) = 59 (23%), and steatosis prevalence of grade 0 = 169 (65%), grade 1 = 64(25%), grade 2/3 = 27(10%). Median HOMA-IR was 1.8 (interquartile range: 1.2-2.7); 100 (38%) patients had HOMA-IR >2. Factors independently associated with significant fibrosis included HOMA-IR (odds ratio [OR] = 8.42), necro-inflammatory grade (OR = 3.17), age (OR = 1.07) and serum total cholesterol levels (OR = 0.008). This was similar to non-Asian region patients, but steatosis was not associated with significant fibrosis in either cohort.Conclusions: In this subgroup study of Asian-region HCV genotype 2 or 3 patients, insulin resistance, along with age, cholesterol levels and necro-inflammation, but not steatosis may be associated with significant hepatic fibrosis.

BACKGROUND & AIMS: It is recommended that patients with chronic hepatitis C virus (HCV) genotype 3 infections receive 24 weeks of treatment. A rapid virologic response (RVR; at week 4) predicts a sustained virologic response (SVR), although not all patients with an RVR achieve an SVR. We explored the relationships among hepatic steatosis, level of HCV RNA, relapse, and RVR in a phase 3 randomized controlled trial of 932 patients infected with HCV genotype 2 (n = 427) or 3 (n = 505) who received 24 weeks of therapy with interferon-alpha. METHODS: In patients with an RVR (HCV RNA <3 IU/mL), the presence of an SVR was modeled using multivariate logistic regression as a function of age, sex, weight, body mass index, insulin resistance, steatosis, and levels of gamma-glutamyl transpeptidase, alanine aminotransferase, liver fibrosis, and baseline HCV RNA. RESULTS: RVR, SVR, and relapse rates among patients with HCV genotype 3 were 79.6%, 79.2%, and 15.6%, respectively; corresponding rates among patients with HCV genotype 2 were 86.7%, 84.3%, and 10.1%. An RVR had high predictive value for an SVR in patients with HCV genotypes 2 (88.9%) and 3 (88.1%). The strongest independent predictors of relapse in patients with genotype 3 and an RVR were steatosis (odds ratio 3.0; P = .003) and HCV RNA >= 400,000 IU/mL (odds ratio 2.5; P = .04). Relapse rates in patients with steatosis were 17.4% and 20.9% for low and high baseline levels of HCV RNA, respectively; corresponding rates in those without steatosis were 2.5% and 8.8%. CONCLUSIONS: Steatosis was associated with significantly higher rates of relapse, irrespective of viral load, in patients infected with HCV genotype 3 who had an RVR. Further studies are needed to determine if longer treatment durations are effective in patients with an RVR and these risk factors.

While 20–40% of patients with hepatitis C virus (HCV) monoinfection will spontaneously clear the virus, less is known regarding clearance with coinfections. HCV, human immunodeficiency virus (HIV), and human T-cell lymphotrophic virus 1 and 2 (HTLV-1/2) coinfection occurs due to shared routes of transmission and is prevalent in Brazil.To compare the proportion of patients who have spontaneously cleared HCV in patients with HCV monoinfection to patients coinfected by HCV/HIV, or HCV/HIV/HTLV-1.Using medical records from two clinics in Salvador, Brazil, including demographic data and serological markers of HCV, HIV and HTLV-I/II, cross-sectional data was obtained from 197 patients. Patients who were anti-HCV positive and HCV RNA negative, and who did not receive HCV treatment were defined as having cleared infection.Nineteen patients (9.5%) showed evidence of spontaneous HCV clearance; with clearance in 9 of 108 (8.3%) patients in the HCV monoinfected group, 5 of 68 (7.4%) patients with HCV/HIV, and 5 of 21 (23.8%) patients with HCV/HIV/HTLV. Demographic data were not associated with HCV clearance status. Patients coinfected with both HIV and HTLV-1 had increased odds (5.50; 95% CI 1.00, 30.17) of spontaneous clearance of HCV compared with patients who were HIV negative or of unknown HIV status.Our study found that patients coinfected with HIV and HTLV-1 were more likely to spontaneously clear hepatitis C virus than patients with HIV/HCV or HCV alone. The effects of HTLV coinfection on the immune response of such patients may be associated with these findings.

Methods: Results from two Phase 1 studies were pooled for the analysis. Study 1 investigated the antiviral activity of oral filibuvir doses of 100, 300 and 450mg BID and 300mg TID or placebo, for 8 days in treatment naive patients. In the second study, the antiviral activity of filibuvir 450mg BID in treatment experienced patients dosed for 10 days and 700mg BID in treatment naive patients dosed for 3 days was evaluated. Filibuvir exposures achieved over 24 hours (AUC24) were utilized to inform the ER analysis of the maximum log change in viral load from baseline. The relationship was described by an inhibitory Emax model using a non-linear mixed effects model. Effects of food, baseline viral load (BVL) and genotype (1a/1b) on relevant model parameters were also evaluated. The antiviral response for a range of doses was simulated from the model. Results: The ER data from 52 subjects were adequately described by the Emax model with Emax (95%CI) = −2.14 (−2.54, −2.02) IU/ml and AUC2450 (95%CI) = 54,176 (44,606, 102,006) ng·hr/ml. BVL was the only influential covariate which described the maximal response (Emax). The analysis indicated that doses in excess of 300mg BID were expected to achieve atleast 70% of Emax. Furthermore, filibuvir exposures resulting from doses greater than 600mg BID produced responses close to Emax indicating that higher doses (>600mg BID) are unlikely to provide additional efficacy. Conclusions: Filibuvir exposures adequately described the maximum log change in viral load relative to baseline. The above analysis in conjunction with results from a Phase 2a study (200/300/500mg BID+SOC, rapid viral response rates of 60–75%) suggests that filibuvir doses of 300 and 600mg BID are expected to be efficacious when combined with SOC and will be further investigated in Phase 2b studies.

Methods: Results from two Phase 1 studies were pooled for the analysis. Study 1 investigated the antiviral activity of oral filibuvir doses of 100, 300 and 450mg BID and 300mg TID or placebo, for 8 days in treatment naive patients. In the second study, the antiviral activity of filibuvir 450mg BID in treatment experienced patients dosed for 10 days and 700mg BID in treatment naive patients dosed for 3 days was evaluated. Filibuvir exposures achieved over 24 hours (AUC24) were utilized to inform the ER analysis of the maximum log change in viral load from baseline. The relationship was described by an inhibitory Emax model using a non-linear mixed effects model. Effects of food, baseline viral load (BVL) and genotype (1a/1b) on relevant model parameters were also evaluated. The antiviral response for a range of doses was simulated from the model. Results: The ER data from 52 subjects were adequately described by the Emax model with Emax (95%CI) = −2.14 (−2.54, −2.02) IU/ml and AUC2450 (95%CI) = 54,176 (44,606, 102,006) ng·hr/ml. BVL was the only influential covariate which described the maximal response (Emax). The analysis indicated that doses in excess of 300mg BID were expected to achieve atleast 70% of Emax. Furthermore, filibuvir exposures resulting from doses greater than 600mg BID produced responses close to Emax indicating that higher doses (>600mg BID) are unlikely to provide additional efficacy. Conclusions: Filibuvir exposures adequately described the maximum log change in viral load relative to baseline. The above analysis in conjunction with results from a Phase 2a study (200/300/500mg BID+SOC, rapid viral response rates of 60–75%) suggests that filibuvir doses of 300 and 600mg BID are expected to be efficacious when combined with SOC and will be further investigated in Phase 2b studies.