A predictive enrichment procedure to identify potential responders to a new therapy for randomized, comparative controlled clinical studies

An explainable machine learning-based phenomapping strategy for adaptive predictive enrichment in randomized clinical trials

Randomized clinical trials (RCT) represent the cornerstone of evidence-based medicine but are resource-intensive. We propose and evaluate a machine learning (ML) strategy of adaptive predictive enrichment through computational trial phenomaps to optimize RCT enrollment. In simulated group sequential analyses of two large cardiovascular outcomes RCTs of (1) a therapeutic drug (pioglitazone versus placebo; Insulin Resistance Intervention after Stroke (IRIS) trial), and (2) a disease management strategy (intensive versus standard systolic blood pressure reduction in the Systolic Blood Pressure Intervention Trial (SPRINT)), we constructed dynamic phenotypic representations to infer response profiles during interim analyses and examined their association with study outcomes. Across three interim timepoints, our strategy learned dynamic phenotypic signatures predictive of individualized cardiovascular benefit. By conditioning a prospective candidate's probability of enrollment on their predicted benefit, we estimate that our approach would have enabled a reduction in the final trial size across ten simulations (IRIS: -14.8% ± 3.1%, pone-sample t-test = 0.001; SPRINT: -17.6% ± 3.6%, pone-sample t-test < 0.001), while preserving the original average treatment effect (IRIS: hazard ratio of 0.73 ± 0.01 for pioglitazone vs placebo, vs 0.76 in the original trial; SPRINT: hazard ratio of 0.72 ± 0.01 for intensive vs standard systolic blood pressure, vs 0.75 in the original trial; all simulations with Cox regression-derived p value of  < 0.01 for the effect of the intervention on the respective primary outcome). This adaptive framework has the potential to maximize RCT enrollment efficiency.

An explainable machine learning-based phenomapping strategy for adaptive predictive enrichment in randomized controlled trials

Randomized controlled trials (RCT) represent the cornerstone of evidence-based medicine but are resource-intensive. We propose and evaluate a machine learning (ML) strategy of adaptive predictive enrichment through computational trial phenomaps to optimize RCT enrollment. In simulated group sequential analyses of two large cardiovascular outcomes RCTs of (1) a therapeutic drug (pioglitazone versus placebo; Insulin Resistance Intervention after Stroke (IRIS) trial), and (2) a disease management strategy (intensive versus standard systolic blood pressure reduction in the Systolic Blood Pressure Intervention Trial (SPRINT)), we constructed dynamic phenotypic representations to infer response profiles during interim analyses and examined their association with study outcomes. Across three interim timepoints, our strategy learned dynamic phenotypic signatures predictive of individualized cardiovascular benefit. By conditioning a prospective candidate's probability of enrollment on their predicted benefit, we estimate that our approach would have enabled a reduction in the final trial size across ten simulations (IRIS: -14.8% ± 3.1%, pone-sample t-test=0.001; SPRINT: -17.6% ± 3.6%, pone-sample t-test<0.001), while preserving the original average treatment effect (IRIS: hazard ratio of 0.73 ± 0.01 for pioglitazone vs placebo, vs 0.76 in the original trial; SPRINT: hazard ratio of 0.72 ± 0.01 for intensive vs standard systolic blood pressure, vs 0.75 in the original trial; all with pone-sample t-test<0.01). This adaptive framework has the potential to maximize RCT enrollment efficiency.

IBIS: identify biomarker-based subgroups with a Bayesian enrichment design for targeted combination therapy

BACKGROUND

Combination therapies directed at multiple targets have potentially improved treatment effects for cancer patients. Compared to monotherapy, targeted combination therapy leads to an increasing number of subgroups and complicated biomarker-based efficacy profiles, making it more difficult for efficacy evaluation in clinical trials. Therefore, it is necessary to develop innovative clinical trial designs to explore the efficacy of targeted combination therapy in different subgroups and identify patients who are more likely to benefit from the investigational combination therapy.

METHODS

We propose a statistical tool called 'IBIS' to Identify BIomarker-based Subgroups and apply it to the enrichment design framework. The IBIS contains three main elements: subgroup division, efficacy evaluation and subgroup identification. We first enumerate all possible subgroup divisions based on biomarker levels. Then, Jensen-Shannon divergence is used to distinguish high-efficacy and low-efficacy subgroups, and Bayesian hierarchical model (BHM) is employed to borrow information within these two subsets for efficacy evaluation. Regarding subgroup identification, a hypothesis testing framework based on Bayes factors is constructed. This framework also plays a key role in go/no-go decisions and enriching specific population. Simulation studies are conducted to evaluate the proposed method.

RESULTS

The accuracy and precision of IBIS could reach a desired level in terms of estimation performance. In regard to subgroup identification and population enrichment, the proposed IBIS has superior and robust characteristics compared with traditional methods. An example of how to obtain design parameters for an adaptive enrichment design under the IBIS framework is also provided.

CONCLUSIONS

IBIS has the potential to be a useful tool for biomarker-based subgroup identification and population enrichment in clinical trials of targeted combination therapy.

A two-stage adaptive clinical trial design with data-driven subgroup identification at interim analysis

In this paper, we consider randomized controlled clinical trials comparing two treatments in efficacy assessment using a time to event outcome. We assume a relatively small number of candidate biomarkers available in the beginning of the trial, which may help define an efficacy subgroup which shows differential treatment effect. The efficacy subgroup is to be defined by one or two biomarkers and cut-offs that are unknown to the investigator and must be learned from the data. We propose a two-stage adaptive design with a pre-planned interim analysis and a final analysis. At the interim, several subgroup-finding algorithms are evaluated to search for a subgroup with enhanced survival for treated versus placebo. Conditional powers computed based on the subgroup and the overall population are used to make decision at the interim to terminate the study for futility, continue the study as planned, or conduct sample size recalculation for the subgroup or the overall population. At the final analysis, combination tests together with closed testing procedures are used to determine efficacy in the subgroup or the overall population. We conducted simulation studies to compare our proposed procedures with several subgroup-identification methods in terms of a novel utility function and several other measures. This research demonstrated the benefit of incorporating data-driven subgroup selection into adaptive clinical trial designs.

Overall and patient-level comparative effectiveness of dimethyl fumarate and fingolimod: A precision medicine application to the Observatoire Français de la Sclérose en Plaques registry

Background

Comparing real-world effectiveness and tolerability of therapies for relapsing-remitting multiple sclerosis is increasingly important, though average treatment effects fail to capture possible treatment effect heterogeneity. With the clinical course of the disease being highly heterogeneous across patients, precision medicine methods enable treatment response heterogeneity investigations.

Objective

To compare real-world effectiveness and discontinuation profiles between dimethyl fumarate and fingolimod while investigating treatment effect heterogeneity with precision medicine methods.

Methods

Adults initiating dimethyl fumarate or fingolimod as a second-line therapy were selected from a French registry. The primary outcome was annualized relapse rate at 12 months. Seven secondary outcomes relative to discontinuation and disease progression were considered. A precision medicine framework was used to characterize treatment effect heterogeneity.

Results

Annualized relapse rates at 12 months were similar for dimethyl fumarate and fingolimod. The odd of treatment persistence was 47% lower for patients treated with dimethyl fumarate relative to those treated with fingolimod (odds ratio: 0.53, 95% confidence interval: 0.39, 0.70). None of the five precision medicine scoring approaches identified treatment heterogeneity.

Conclusion

These findings substantiated the similar effectiveness and different discontinuation profiles for dimethyl fumarate and fingolimod as a second-line therapy for relapsing-remitting multiple sclerosis, with no significant effect heterogeneity observed.

Bayesian two-stage sequential enrichment design for biomarker-guided phase II trials for anticancer therapies

Biomarker-guided phase II trials have become increasingly important for personalized cancer treatment. In this paper, we propose a Bayesian two-stage sequential enrichment design for such biomarker-guided trials. We assumed that all patients were dichotomized as marker positive or marker negative based on their biomarker status; the positive patients were considered more likely to respond to the targeted drug. Early stopping rules and adaptive randomization methods were embedded in the design to control the number of patients receiving inferior treatment. At the same time, a Bayesian hierarchical model was used to borrow information between the positive and negative control arms to improve efficiency. Simulation results showed that the proposed design achieved higher empirical power while controlling the type I error and assigned more patients to the superior treatment arms. The operating characteristics suggested that the design has good performance and may be useful for biomarker-guided phase II trials for evaluating anticancer therapies.

Harrell Jr F, Moons KGM, Debray TPA, Reitsma JB. A tutorial on individualized treatment effect prediction from randomized trials with a binary endpoint

Randomized trials typically estimate average relative treatment effects, but decisions on the benefit of a treatment are possibly better informed by more individualized predictions of the absolute treatment effect. In case of a binary outcome, these predictions of absolute individualized treatment effect require knowledge of the individual's risk without treatment and incorporation of a possibly differential treatment effect (ie, varying with patient characteristics). In this article, we lay out the causal structure of individualized treatment effect in terms of potential outcomes and describe the required assumptions that underlie a causal interpretation of its prediction. Subsequently, we describe regression models and model estimation techniques that can be used to move from average to more individualized treatment effect predictions. We focus mainly on logistic regression-based methods that are both well-known and naturally provide the required probabilistic estimates. We incorporate key components from both causal inference and prediction research to arrive at individualized treatment effect predictions. While the separate components are well known, their successful amalgamation is very much an ongoing field of research. We cut the problem down to its essentials in the setting of a randomized trial, discuss the importance of a clear definition of the estimand of interest, provide insight into the required assumptions, and give guidance with respect to modeling and estimation options. Simulated data illustrate the potential of different modeling options across scenarios that vary both average treatment effect and treatment effect heterogeneity. Two applied examples illustrate individualized treatment effect prediction in randomized trial data.

Incorporating covariates information in adaptive clinical trials for precision medicine

Precision medicine is the systematic use of information that pertains to an individual patient to select or optimize that patient's preventative and therapeutic care. Recent studies have classified biomarkers into predictive and prognostic biomarkers based on their roles in clinical studies. To design a clinical trial for precision medicine, predictive biomarkers and prognostic biomarkers should both be included. In statistical analysis, biomarkers are mathematically treated as covariates. We first classify covariates into predictive and prognostic covariates according to their roles. We then provide a brief review of recent advances in adaptive designs that incorporate covariates. However, the literature includes no designs that incorporate both prognostic covariates and predictive covariates simultaneously. In this paper, we propose a new family of covariate-adjusted response-adaptive (CARA) designs that incorporate both prognostic and predictive covariates and the responses. It is important to note that the predictive biomarkers and prognostic biomarkers play different roles in the new designs. The advantages of the proposed methods are demonstrated via numerical studies, and some further statistical issues are also discussed.

Difference in background factors between responders to gabapentin enacarbil treatment and responders to placebo: pooled analyses of two randomized, double-blind, placebo-controlled studies in Japanese patients with restless legs syndrome

OBJECTIVE

Restless legs syndrome (RLS) is a sensorimotor disorder that is characterized by uncomfortable and unpleasant sensations mainly in the legs. Two placebo-controlled studies (Phase II/III and post-marketing) in Japanese patients with RLS failed to demonstrate the efficacy of gabapentin enacarbil (GE) 600 mg in the change from baseline in International Restless Legs Syndrome Rating Scale (IRLS) score at the end of the treatment period. The high response to placebo is thought to be a possible reason why the post-marketing study failed. The objectives of these post hoc analyses were to determine potential predictive factors associated with improvement in IRLS score with GE treatment and to identify subgroups with higher placebo responses.

METHODS

We combined data from the two Japanese studies and analyzed change from baseline in IRLS score in the pooled population and subgroups defined by several patient characteristics. Moreover, we calculated the variable importance of each factor and performed predictive enrichment analysis to identify an enrichable subpopulation with greater improvement by GE treatment.

RESULTS

The post hoc analyses suggested that higher baseline IRLS score (≥21) and higher body mass index (≥25 kg/m²) were associated with higher placebo responses. On the other hand, positive family history of RLS, prior use of dopaminergic receptor agonists, and higher baseline ferritin level (≥50 ng/mL) were associated with higher responses to GE.

CONCLUSIONS

Our results suggest that patients with typical idiopathic RLS characteristics, including positive family history and no low ferritin level, would be expected to derive the greatest benefits from GE treatment.

Learning and confirming a class of treatment responders in clinical trials

Clinical trials require substantial effort and time to complete, and regulatory agencies may require two successful efficacy trials before approving a new drug. One way to improve the chance of follow-up success is to identify a subpopulation among whom treatment effects are estimated to be beneficial, and enrolling future studies from this subpopulation. In this article we study confirmable responder class (CRC) learning, where the objective is to learn in a random half of the dataset (training set) a subpopulation among whom the predicted conditional ATE (CATE) suggests clinically meaningful benefit, with maximum power of being confirmed via hypothesis test in the other half (test set). We studied a set of CRC learners across simulated datasets in which either all patients benefited, or only some did. Performance metrics included the rate of confirmation in the test set, and the classification accuracy of the CRC compared with the group with true CATE>0. In trials where all patients benefitted, only two learners were able to consistently identify most of the population as the CRC. One of these also performed especially well when only some patients benefitted, having relatively high confirmation rates, and showing robustness to the dimension of the covariate vector and population characteristics. This learner is based on cross-validation, and is a possible avenue for further development of model selection criteria for CRC learning. We also show that the performance of all methods can be improved by using both halves of the original dataset as training and test sets in turn.

A scoping review of causal methods enabling predictions under hypothetical interventions

BACKGROUND

The methods with which prediction models are usually developed mean that neither the parameters nor the predictions should be interpreted causally. For many applications, this is perfectly acceptable. However, when prediction models are used to support decision making, there is often a need for predicting outcomes under hypothetical interventions.

AIMS

We aimed to identify published methods for developing and validating prediction models that enable risk estimation of outcomes under hypothetical interventions, utilizing causal inference. We aimed to identify the main methodological approaches, their underlying assumptions, targeted estimands, and potential pitfalls and challenges with using the method. Finally, we aimed to highlight unresolved methodological challenges.

METHODS

We systematically reviewed literature published by December 2019, considering papers in the health domain that used causal considerations to enable prediction models to be used for predictions under hypothetical interventions. We included both methodologies proposed in statistical/machine learning literature and methodologies used in applied studies.

RESULTS

We identified 4919 papers through database searches and a further 115 papers through manual searches. Of these, 87 papers were retained for full-text screening, of which 13 were selected for inclusion. We found papers from both the statistical and the machine learning literature. Most of the identified methods for causal inference from observational data were based on marginal structural models and g-estimation.

CONCLUSIONS

There exist two broad methodological approaches for allowing prediction under hypothetical intervention into clinical prediction models: (1) enriching prediction models derived from observational studies with estimated causal effects from clinical trials and meta-analyses and (2) estimating prediction models and causal effects directly from observational data. These methods require extending to dynamic treatment regimes, and consideration of multiple interventions to operationalise a clinical decision support system. Techniques for validating 'causal prediction models' are still in their infancy.

Counterfactual clinical prediction models could help to infer individualized treatment effects in randomized controlled trials-An illustration with the International Stroke Trial

OBJECTIVE

Causal treatment effects are estimated at the population level in randomized controlled trials, while clinical decision is often to be made at the individual level in practice. We aim to show how clinical prediction models used under a counterfactual framework may help to infer individualized treatment effects.

STUDY DESIGN AND SETTING

As an illustrative example, we reanalyze the International Stroke Trial. This large, multicenter trial enrolled 19,435 adult patients with suspected acute ischemic stroke from 36 countries, and reported a modest average benefit of aspirin (vs. no aspirin) on a composite outcome of death or dependency at 6 months. We derive and validate multivariable logistic regression models that predict the patient counterfactual risks of outcome with and without aspirin, conditionally on 23 predictors.

RESULTS

The counterfactual prediction models display good performance in terms of calibration and discrimination (validation c-statistics: 0.798 and 0.794). Comparing the counterfactual predicted risks on an absolute difference scale, we show that aspirin-despite an average benefit-may increase the risk of death or dependency at 6 months (compared with the control) in a quarter of stroke patients.

CONCLUSIONS

Counterfactual prediction models could help researchers and clinicians (i) infer individualized treatment effects and (ii) better target patients who may benefit from treatments.

A Novel DBS Paradigm for Axial Features in Parkinson's Disease: A Randomized Crossover Study

BACKGROUND

High-frequency (130-185 Hz) deep brain stimulation (DBS) of the subthalamic nucleus is more effective for appendicular than axial symptoms in Parkinson's disease (PD). Low-frequency (60-80 Hz) stimulation (LFS) may reduce gait/balance impairment but typically results in worsening appendicular symptoms. We created a "dual-frequency" programming paradigm (interleave-interlink, IL-IL) to address both axial and appendicular symptoms. In IL-IL, 2 overlapping LFS programs are applied to the DBS lead, with the overlapping area focused on the optimal cathode. The nonoverlapping area (LFS) is thought to reduce gait/balance impairment, whereas the overlapping area (high-frequency stimulation, HFS) aims to control appendicular symptoms.

METHODS

We performed a randomized, double-blind crossover trial comparing patients' previously optimized IL-IL and conventional HFS paradigms. Each arm was 2 weeks in duration. The primary outcome measure was the patient/caregiver Modified Clinical Global Impression Severity (CGI-S). Secondary outcome measures included blinded motor evaluations, timed tests, patient/caregiver questionnaires, and Personal KinetiGraphs (PKG).

RESULTS

Twenty-five patients were enrolled, and 20 completed. The patient/caregiver CGI-S for gait/balance (P = 0.01) and appendicular symptom control (P = 0.001), and the blinded rater MDS-UPDRS-III (-5.22, P = 0.02), CGI-S gait/balance (P = 0.01), and CGI-S speech (P = 0.02) were better while on IL-IL. Scores on Parkinson's Disease Quality of Life (P = 0.002) and Freezing-of-Gait Questionnaires (P = 0.04) were better on IL-IL. The Timed-Up-and-Go was 9.8% faster (P = 0.01), with 11.8% reduction in steps (P = 0.001) on IL-IL. There was no difference in PKG bradykinesia (P = 0.18) or tremor (P = 0.23) between paradigms.

CONCLUSIONS

Our results prompt consideration of this novel programming paradigm (IL-IL) for PD patients with axial symptom impairment as a new treatment option for both axial and appendicular symptoms. © 2020 International Parkinson and Movement Disorder Society.

Addressing heterogeneity in amyotrophic lateral sclerosis CLINICAL TRIALS

Amyotrophic lateral sclerosis (ALS) is a debilitating neurodegenerative disorder with complex biology and significant clinical heterogeneity. Many preclinical and early phase ALS clinical trials have yielded promising results that could not be replicated in larger phase 3 confirmatory trials. One reason for the lack of reproducibility may be ALS biological and clinical heterogeneity. Therefore, in this review, we explore sources of ALS heterogeneity that may reduce statistical power to evaluate efficacy in ALS trials. We also review efforts to manage clinical heterogeneity, including use of validated disease outcome measures, predictive biomarkers of disease progression, and individual clinical risk stratification. We propose that personalized prognostic models with use of predictive biomarkers may identify patients with ALS for whom a specific therapeutic strategy may be expected to be more successful. Finally, the rapid application of emerging clinical and biomarker strategies may reduce heterogeneity, increase trial efficiency, and, in turn, accelerate ALS drug development.

A Th1/IFNγ Gene Signature Is Prognostic in the Adjuvant Setting of Resectable High-Risk Melanoma but Not in Non-Small Cell Lung Cancer

PURPOSE

Immune components of the tumor microenvironment (TME) have been associated with disease outcome. We prospectively evaluated the association of an immune-related gene signature (GS) with clinical outcome in melanoma and non-small cell lung cancer (NSCLC) tumor samples from two phase III studies.

EXPERIMENTAL DESIGN

The GS was prospectively validated using an adaptive signature design to optimize it for the sample type and technology used in phase III studies. One-third of the samples were used as "training set"; the remaining two thirds, constituting the "test set," were used for the prospective validation of the GS.

RESULTS

In the melanoma training set, the expression level of eight Th1/IFNγ-related genes in tumor-positive lymph node tissue predicted the duration of disease-free survival (DFS) and overall survival (OS) in the placebo arm. This GS was prospectively and independently validated as prognostic in the test set. Building a multivariate Cox model in the test set placebo patients from clinical covariates and the GS score, an increased number of melanoma-involved lymph nodes and the GS were associated with DFS and OS. This GS was not associated with DFS in NSCLC, although expression of the Th1/IFNγ-related genes was associated with the presence of lymphocytes in tumor samples in both indications.

CONCLUSIONS

These findings provide evidence that expression of Th1/IFNγ genes in the TME, as measured by this GS, is associated with clinical outcome in melanoma. This suggests that, using this GS, patients with stage IIIB/C melanoma can be classified into different risk groups.

Meta-Enrichment Analyses to Identify Advanced Gastric Cancer Patients Who Achieve a Higher Response to S-1/Cisplatin

The Multicenter phase III comparison of cisplatin/S-1 with cisplatin/infusional fluorouracil in advanced gastric or gastroesophageal adenocarcinoma study (FLAGS) and the Diffuse Gastric and Esophagogastric Junction Cancer S-1 Trial (DIGEST) have shown that patients with advanced gastric cancer treated with S-1/Cisplatin (CS) have similar overall survival (OS) compared to 5-fluorouracil/cisplatin (CF). The purpose of this analysis was to identify patients who may specifically benefit from CS using meta-enrichment analysis of the combined two datasets. Eleven clinico-pathological factors were selected and a high response enrichable population was determined. The efficacy of CS in the combined data set of 1365 patients (n = 1019 from FLAGS and n = 346 from DIGEST) was analyzed. We identified 683 patients (n = 374 from CS, n = 309 from CF) as the high response enrichable population who were classified as those with Eastern Cooperative Oncology Group Performance Status (ECOG PS) 1, more than two metastatic sites and low neutrophil-lymphocyte ratio (log(NL ratio)). In the high response enrichable population, the median OS in the CS group was 241 days compared to 210 days in the CF group (hazard ratio 0.776; 95% confidence interval 0.658 to 0.915; p-value 0.004). Through meta-enrichment analysis, the high response enrichable population to CS was identified. Our findings show the clinical importance of selecting the appropriate treatment based on specific patient characteristics.

A proof-of-concept application of a novel scoring approach for personalized medicine in multiple sclerosis

BACKGROUND

Stratified medicine methodologies based on subgroup analyses are often insufficiently powered. More powerful personalized medicine approaches are based on continuous scores.

OBJECTIVE

We deployed a patient-specific continuous score predicting treatment response in patients with relapsing-remitting multiple sclerosis (RRMS).

METHODS

Data from two independent randomized controlled trials (RCTs) were used to build and validate an individual treatment response (ITR) score, regressing annualized relapse rates (ARRs) on a set of baseline predictors.

RESULTS

The ITR score for the combined treatment groups versus placebo detected differential clinical response in both RCTs. High responders in one RCT had a cross-validated ARR ratio of 0.29 (95% confidence interval (CI) = 0.13-0.55) versus 0.62 (95% CI = 0.47-0.83) for all other responders (heterogeneity p = 0.038) and were validated in the other RCT, with the corresponding ARR ratios of 0.31 (95% CI = 0.18-0.56) and 0.61 (95% CI = 0.47-0.79; heterogeneity p = 0.036). The strongest treatment effect modifiers were the Short Form-36 Physical Component Summary, age, Visual Function Test 2.5%, prior MS treatment and Expanded Disability Status Scale.

CONCLUSION

Our modelling strategy detects and validates an ITR score and opens up avenues for building treatment response calculators that are also applicable in routine clinical practice.

Interpretability of Cancer Clinical Trial Results Using Restricted Mean Survival Time as an Alternative to the Hazard Ratio

Importance

In a comparative clinical study with progression-free survival (PFS) or overall survival (OS) as the end point, the hazard ratio (HR) is routinely used to design the study and to estimate the treatment effect at the end of the study. The clinical interpretation of the HR may not be straightforward, especially when the underlying model assumption is not valid. A robust procedure for study design and analysis that enables clinically meaningful interpretation of trial results is warranted.

Objective

To discuss issues of conventional trial design and analysis and to present alternatives to the HR using a recent immunotherapy study as an illustrative example.

Design, Setting, and Participants

By comparing 2 groups in a survival analysis, we discuss issues of using the HR and present the restricted mean survival time (RMST) as a summary measure of patients’ survival profile over time. We show how to use the difference or ratio in RMST between 2 groups as an alternative for designing and analyzing a clinical study with an immunotherapy study as an illustrative example.

Main Outcomes and Measures

Overall survival or PFS. Group contrast measures included HR, RMST difference or ratio, and the event rate difference.

Results

For the illustrative example, the HR procedure indicates that nivolumab significantly prolonged patient OS and was numerically better than docetaxel for PFS. However, the median PFS time of docetaxel was significantly better than that of nivolumab. Therefore, it may be difficult to use median OS and/or PFS to interpret of the HR value clinically. On the other hand, using RMST difference, nivolumab was significantly better than docetaxel for both OS and PFS. We also provide details regarding design of a future study with RMST-based measures.

Conclusions and Relevance

The design and analysis of a conventional cancer clinical trial can be improved by adopting a robust statistical procedure that enables clinically meaningful interpretations of the treatment effect. The RMST-based quantitative method may be used as a primary tool for future cancer trials or to help us to better understand the clinical interpretation of the HR even when its model assumption is plausible.

Acute Respiratory Distress Syndrome Phenotypes

The acute respiratory distress syndrome (ARDS) phenotype was first described over 50 years ago and since that time significant progress has been made in understanding the biologic processes underlying the syndrome. Despite this improved understanding, no pharmacologic therapies aimed at the underlying biology have been proven effective in ARDS. Increasingly, ARDS has been recognized as a heterogeneous syndrome characterized by subphenotypes with distinct clinical, radiographic, and biologic differences, distinct outcomes, and potentially distinct responses to therapy. The Berlin Definition of ARDS specifies three severity classifications: mild, moderate, and severe based on the PaO₂ to FiO₂ ratio. Two randomized controlled trials have demonstrated a potential benefit to prone positioning and neuromuscular blockade in moderate to severe phenotypes of ARDS only. Precipitating risk factor, direct versus indirect lung injury, and timing of ARDS onset can determine other clinical phenotypes of ARDS after admission. Radiographic phenotypes of ARDS have been described based on a diffuse versus focal pattern of infiltrates on chest imaging. Finally and most promisingly, biologic subphenotypes or endotypes have increasingly been identified using plasma biomarkers, genetics, and unbiased approaches such as latent class analysis. The potential of precision medicine lies in identifying novel therapeutics aimed at ARDS biology and the subpopulation within ARDS most likely to respond. In this review, we discuss the challenges and approaches to subphenotype ARDS into clinical, radiologic, severity, and biologic phenotypes with an eye toward the future of precision medicine in critical care.

Subgroup identification in clinical trials via the predicted individual treatment effect

Identifying subgroups of treatment responders through the different phases of clinical trials has the potential to increase success in drug development. Recent developments in subgroup analysis consider subgroups that are defined in terms of the predicted individual treatment effect, i.e. the difference between the predicted outcome under treatment and the predicted outcome under control for each individual, which in turn may depend on multiple biomarkers. In this work, we study the properties of different modelling strategies to estimate the predicted individual treatment effect. We explore linear models and compare different estimation methods, such as maximum likelihood and the Lasso with and without randomized response. For the latter, we implement confidence intervals based on the selective inference framework to account for the model selection stage. We illustrate the methods in a dataset of a treatment for Alzheimer disease (normal response) and in a dataset of a treatment for prostate cancer (survival outcome). We also evaluate via simulations the performance of using the predicted individual treatment effect to identify subgroups where a novel treatment leads to better outcomes compared to a control treatment.

A single-index threshold Cox proportional hazard model for identifying a treatment-sensitive subset based on multiple biomarkers

In this paper, we introduce a single-index threshold Cox proportional hazard model to select and combine biomarkers to identify patients who may be sensitive to a specific treatment. A penalized smoothed partial likelihood is proposed to estimate the parameters in the model. A simple, efficient, and unified algorithm is presented to maximize this likelihood function. The estimators based on this likelihood function are shown to be consistent and asymptotically normal. Under mild conditions, the proposed estimators also achieve the oracle property. The proposed approach is evaluated through simulation analyses and application to the analysis of data from two clinical trials, one involving patients with locally advanced or metastatic pancreatic cancer and one involving patients with resectable lung cancer.

Identifying treatment responders using counterfactual modeling and potential outcomes

Individualizing treatment according to patients' characteristics is central for personalized or precision medicine. There has been considerable recent research in developing statistical methods to determine optimal personalized treatment strategies by modeling the outcome of patients according to relevant covariates under each of the alternative treatments, and then relying on so-called predicted individual treatment effects. In this paper, we use potential outcomes and principal stratification frameworks and develop a multinomial model for left and right-censored data to estimate the probability that a patient is a responder given a set of baseline covariates. The model can apply to RCT or observational study data. This method is based on the monotonicity assumption, which implies that no patients would respond to the control treatment but not to the experimental one. We conduct a simulation study to evaluate the properties of the proposed estimation method. Results showed that the predictions of the probability of being a responder were well calibrated even if we observed variability and a small bias when many parameters were estimated. We finally applied the method to a cohort study on the selection of patients for additional radiotherapy after resection of a soft-tissue sarcoma.

Identifying a subpopulation with higher likelihoods of early response to treatment in a heterogeneous rare disease: a post hoc study of response to teduglutide for short bowel syndrome

Purpose

Teduglutide, a glucagon-like peptide-2 analog, has demonstrated efficacy in reducing parenteral support (PS) among patients with short bowel syndrome with intestinal failure (SBS–IF). This study aims to identify a subpopulation of SBS–IF patients for whom teduglutide has an especially pronounced effect.

Patients and methods

Data were from a 24-week, Phase III trial (Study of Teduglutide Effectiveness in Parenteral Nutrition-Dependent SBS Subjects; NCT00798967) that randomized SBS–IF patients with PS dependency to receive teduglutide (n=43) or placebo (n=43). Two prediction models (1 for each arm) were developed for response, defined as 20% reduction in weekly PS at Weeks 20 and 24. Potential predictors included demographics, disease characteristics, and concomitant medications. Patients were then ranked based on the effect score, an individualized predicted response rate difference with teduglutide versus placebo. A subpopulation of patients with a pronounced benefit from teduglutide versus placebo was identified. Baseline characteristics and clinical outcomes were compared between patients included versus those not included in the subpopulation.

Results

Six predictors of response to teduglutide were selected: older age, volvulus as the cause of major intestinal resection, baseline PS volume >6 L per week, longer time since start of PS dependency, absence of ileocecal valve, and lower percentage of colon remaining. Higher percentage of colon remaining and volvulus were the selected predictors for response to placebo. A subpopulation of patients more likely to respond to teduglutide was identified as those with the top 60% effect scores. The difference in response rate between teduglutide and placebo was 62% in the subpopulation, which was substantially higher than the difference of 33% in the overall population. Mean PS day reduction was also significantly higher for teduglutide compared to placebo in the subpopulation.

Conclusion

Pretreatment characteristics as predictors of response to teduglutide versus placebo within 24 weeks were identifiable in the clinical trial population of SBS–IF patients.

MAGE-A3 immunotherapeutic as adjuvant therapy for patients with resected, MAGE-A3-positive, stage III melanoma (DERMA): a double-blind, randomised, placebo-controlled, phase 3 trial

BACKGROUND

Despite newly approved treatments, metastatic melanoma remains a life-threatening condition. We aimed to evaluate the efficacy of the MAGE-A3 immunotherapeutic in patients with stage IIIB or IIIC melanoma in the adjuvant setting.

METHODS

DERMA was a phase 3, double-blind, randomised, placebo-controlled trial done in 31 countries and 263 centres. Eligible patients were 18 years or older and had histologically proven, completely resected, stage IIIB or IIIC, MAGE-A3-positive cutaneous melanoma with macroscopic lymph node involvement and an Eastern Cooperative Oncology Group performance score of 0 or 1. Randomisation and treatment allocation at the investigator sites were done centrally via the internet. We randomly assigned patients (2:1) to receive up to 13 intramuscular injections of recombinant MAGE-A3 with AS15 immunostimulant (MAGE-A3 immunotherapeutic; 300 μg MAGE-A3 antigen plus 420 μg CpG 7909 reconstituted in AS01B to a total volume of 0·5 mL), or placebo, over a 27-month period: five doses at 3-weekly intervals, followed by eight doses at 12-weekly intervals. The co-primary outcomes were disease-free survival in the overall population and in patients with a potentially predictive gene signature (GS-positive) identified previously and validated here via an adaptive signature design. The final analyses included all patients who had received at least one dose of study treatment; analyses for efficacy were in the as-randomised population and for safety were in the as-treated population. This trial is registered with ClinicalTrials.gov, number NCT00796445.

FINDINGS

Between Dec 1, 2008, and Sept 19, 2011, 3914 patients were screened, 1391 randomly assigned, and 1345 started treatment (n=895 for MAGE-A3 and n=450 for placebo). At final analysis (data cutoff May 23, 2013), median follow-up was 28·0 months [IQR 23·3-35·5] in the MAGE-A3 group and 28·1 months [23·7-36·9] in the placebo group. Median disease-free survival was 11·0 months (95% CI 10·0-11·9) in the MAGE-A3 group and 11·2 months (8·6-14·1) in the placebo group (hazard ratio [HR] 1·01, 0·88-1·17, p=0·86). In the GS-positive population, median disease-free survival was 9·9 months (95% CI 5·7-17·6) in the MAGE-A3 group and 11·6 months (5·6-22·3) in the placebo group (HR 1·11, 0·83-1·49, p=0·48). Within the first 31 days of treatment, adverse events of grade 3 or worse were reported by 126 (14%) of 894 patients in the MAGE-A3 group and 56 (12%) of 450 patients in the placebo group, treatment-related adverse events of grade 3 or worse by 36 (4%) patients given MAGE-A3 vs six (1%) patients given placebo, and at least one serious adverse event by 14% of patients in both groups (129 patients given MAGE-A3 and 64 patients given placebo). The most common adverse events of grade 3 or worse were neoplasms (33 [4%] patients in the MAGE-A3 group vs 17 [4%] patients in the placebo group), general disorders and administration site conditions (25 [3%] for MAGE-A3 vs four [<1%] for placebo) and infections and infestations (17 [2%] for MAGE-A3 vs seven [2%] for placebo). No deaths were related to treatment.

INTERPRETATION

An antigen-specific immunotherapeutic alone was not efficacious in this clinical setting. Based on these findings, development of the MAGE-A3 immunotherapeutic for use in melanoma has been stopped.

FUNDING

GlaxoSmithKline Biologicals SA.

S-1 versus placebo in patients with sorafenib-refractory advanced hepatocellular carcinoma (S-CUBE): a randomised, double-blind, multicentre, phase 3 trial

BACKGROUND

Unresectable advanced hepatocellular carcinoma is a heterogeneous disease, for which sorafenib is the first targeted agent approved for first-line therapy, and treatment options for patients with sorafenib-refractory advanced hepatocellular carcinoma are limited. We assessed the efficacy and safety of S-1, a chemotherapeutic agent based on fluorouracil, in patients with sorafenib-refractory advanced hepatocellular carcinoma.

METHODS

We did a randomised, double-blind, placebo-controlled, phase 3 study done at 57 sites in Japan. Patients with advanced hepatocellular carcinoma who were ineligible for surgical or local-regional therapy and judged refractory to sorafenib (ie, had progressed on sorafenib or had discontinued sorafenib because of adverse events) were randomly assigned (2:1) to receive oral S-1 (weight-banded 80 mg/m² [80-120 mg per day]), or placebo, twice per day for 28 days consecutively, followed by a minimum 14 day drug-free period. This cycle was repeated until disease progression or the patient became intolerant to the study treatment. Patients were stratified by site and presence or absence of extrahepatic metastasis or vascular invasion. The primary endpoint was overall survival, assessed in the full analysis set (ie, all patients who were treated with study drug except any individuals who were found not to have hepatocellular carcinoma or who were found to have active double cancer). Patients, medical staff, investigators, and the sponsor were masked to treatment assignment. Blinding was maintained even after study treatment concluded. This study is registered with JapicCTI, number JapicCTI-090920, and has been completed.

FINDINGS

Between Oct 26, 2009, and Aug 22, 2012, we screened 399 patients. 65 patients were excluded due to not meeting criteria (n=61), declining to participate (n=3), or other reasons (n=1). 334 patients were randomly assigned to receive either S-1 (n=223) or placebo (n=111). One patient in the S-1 group did not receive treatment, and was thus excluded from analyses. At data cutoff, median follow-up was 32·4 months (IQR 24·0-34·7) in the S-1 group and 32·9 months (23·7-39·5) in the placebo group. Median overall survival was 11·1 months (95% CI 9·7-13·1) in the S-1 group and 11·2 months (9·2-12·8) in the placebo group (hazard ratio 0·86, 95% CI 0·67-1·10; p=0·220). The most frequently reported adverse events were skin hyperpigmentation (123 [55%] of 222 patients in the S-1 group vs nine [8%] of 111 patients in the placebo group), decreased appetite (104 [47%] vs 21 [19%]), fatigue (102 [46%] vs 20 [18%]), diarrhoea (77 [35%] vs 14 [13%]), and increased blood bilirubin (77 [35%] vs 14 [13%]). Serious adverse events were reported in 90 (41%) of 222 patients in the S-1 group and 24 (22%) of 111 patients in the placebo group. Five treatment-related deaths were reported in the S-1 group.

INTERPRETATION

S-1 did not prolong overall survival in patients with sorafenib-refractory advanced hepatocellular carcinoma. Further research is needed to identify subgroups of patients who might benefit from S-1.

FUNDING

Taiho Pharmaceuticals.

Optimal stratification in outcome prediction using baseline information

A common practice in predictive medicine is to use current study data to construct a stratification procedure, which groups subjects according to baseline information and forms stratum-specific prevention or intervention strategies. A desirable stratification scheme would not only have small intra-stratum variation but also have a clinically meaningful discriminatory capability. We show how to obtain optimal stratification rules with such desirable properties from fitting a set of regression models relating the outcome to baseline covariates and creating scoring systems for predicting potential outcomes. We propose that all available optimal stratifications be evaluated with an independent dataset to select a final stratification. Lastly, we obtain inferential results for this selected stratification scheme with a holdout dataset. When only one study of moderate size is available, we combine the first two steps via crossvalidation. Extensive simulation studies are used to compare the proposed stratification strategy with alternatives. We illustrate the new proposal using an AIDS clinical trial for binary outcomes and a cardiovascular clinical study for censored event time outcomes.

Testing interaction between treatment and high-dimensional covariates in randomized clinical trials

In this paper, we considered different methods to test the interaction between treatment and a potentially large number (p) of covariates in randomized clinical trials. The simplest approach was to fit univariate (marginal) models and to combine the univariate statistics or p-values (e.g., minimum p-value). Another possibility was to reduce the dimension of the covariates using the principal components (PCs) and to test the interaction between treatment and PCs. Finally, we considered the Goeman global test applied to the high-dimensional interaction matrix, adjusted for the main (treatment and covariates) effects. These tests can be used for personalized medicine to test if a large set of biomarkers can be useful to identify a subset of patients who may be more responsive to treatment. We evaluated the performance of these methods on simulated data and we applied them on data from two early phases oncology clinical trials.

Meta-STEPP: subpopulation treatment effect pattern plot for individual patient data meta-analysis

We have developed a method, called Meta-STEPP (subpopulation treatment effect pattern plot for meta-analysis), to explore treatment effect heterogeneity across covariate values in the meta-analysis setting for time-to-event data when the covariate of interest is continuous. Meta-STEPP forms overlapping subpopulations from individual patient data containing similar numbers of events with increasing covariate values, estimates subpopulation treatment effects using standard fixed-effects meta-analysis methodology, displays the estimated subpopulation treatment effect as a function of the covariate values, and provides a statistical test to detect possibly complex treatment-covariate interactions. Simulation studies show that this test has adequate type-I error rate recovery as well as power when reasonable window sizes are chosen. When applied to eight breast cancer trials, Meta-STEPP suggests that chemotherapy is less effective for tumors with high estrogen receptor expression compared with those with low expression. Copyright © 2016 John Wiley & Sons, Ltd.

Statistical Methods for Establishing Personalized Treatment Rules in Oncology

The process for using statistical inference to establish personalized treatment strategies requires specific techniques for data-analysis that optimize the combination of competing therapies with candidate genetic features and characteristics of the patient and disease. A wide variety of methods have been developed. However, heretofore the usefulness of these recent advances has not been fully recognized by the oncology community, and the scope of their applications has not been summarized. In this paper, we provide an overview of statistical methods for establishing optimal treatment rules for personalized medicine and discuss specific examples in various medical contexts with oncology as an emphasis. We also point the reader to statistical software for implementation of the methods when available.