A win ratio-based framework to combine multiple clinical endpoints in exploratory basket trials.

In contemporary exploratory phase of oncology drug development, there has been an increasing interest in evaluating investigational drug or drug combination in multiple tumor indications in a single basket trial to expedite drug development. There has been extensive research on more efficiently borrowing information across tumor indications in early phase drug development including Bayesian hierarchical modeling and the pruning-and-pooling methods. Despite the fact that the Go/No-Go decision for subsequent Phase 2 or Phase 3 trial initiation is almost always a multi-facet consideration, the statistical literature of basket trial design and analysis has largely been limited to a single binary endpoint. In this paper we explore the application of considering clinical priorities of multiple endpoints based on matched win ratio to the basket trial design and analysis. The control arm data will be simulated for each tumor indication based on the corresponding null assumptions that could be heterogeneous across tumor indications. The matched win ratio matching on the tumor indication can be performed for individual tumor indication, pooled data, or the pooled data after pruning depending on whether an individual evaluation or a simple pooling or a pruning-and-pooling method is used. We conduct the simulation studies to evaluate the performance of proposed win ratio-based framework and the results suggest the proposed framework could provide desirable operating characteristics.

Covariate-adjusted value-guided subgroup identification via boosting.

It is widely recognized that treatment effects could differ across subgroups of patients. Subgroup analysis, which assesses such heterogeneity, provides valuable information in developing personalized therapies. There has been extensive research developing novel statistical methods for subgroup identification. The recent contribution is a value-guided subgroup identification method that directly maximizes treatment benefit at the subgroup level for survival outcome, rather than relying on individual treatment effect estimation. In this paper, we first completed this framework by illustrating its application to continuous and binary outcomes. More importantly, we extended the original framework to account for the prognostic effects and named this new method Covariate-Adjusted Value-guided subgroup identification via boosting (CAVboost). The original method directly used the outcome to formulate the value function for subgroup identification. Since the outcome can further be decomposed as prognostic effects and treatment effects, specifying the prognostic effects as the covariates of a model for the outcome can single out the treatment effects and improve the power to detect them across subgroups. Our proposed CAVboost was based on this key idea. It used a covariate-adjusted treatment effect estimator, instead of the outcome itself, to formulate the value function for subgroup identification. CAVboost estimates the treatment effect by using covariates to account for the prognostic effects, which mimics the idea of using covariates in an ANCOVA estimator. We showed that CAVboost could effectively improve the subgroup identification capability for both continuous and binary outcomes.

An extension to a 2-in-1 adaptive design with biomarker subpopulation selection.

In a recent article, Jin and Zhang (2022) proposed an adaptive 2-in-1 design which can expand an ongoing Phase 2 trial with multiple treatment or dose arms into a confirmatory Phase 3 trial with the selected arms based on interim data, and proved that the design can preserve the familywise Type I error rate under a mild assumption. The proposed adaptive design provides an efficient pathway to combine the treatment or dose selection stage and the confirmatory stage into one trial and can expedite drug development. Here we extend the adaptive 2-in-1 design by Jin and Zhang (2022) to an adaptive 2-in-1 design with biomarker subpopulation selection with a similar framework.

An Extended Framework of Multiple Testing in Group Sequential Design.

Here, we consider testing multiple hypotheses in group sequential trials. A graphical multiple test procedure was proposed for group sequential trials using weighted Bonferroni test. In this paper, we extend the framework for the graph-based group sequential procedure by applying a modified weighted Simes test. The proposed procedure preserves the familywise error rate. Simulations are conducted to evaluate the performances of the proposed procedure. The proposed procedures are also illustrated with a numerical example.

A variation of a 2-in-1 adaptive design to seamlessly expand a selected dose from a phase 2 trial to a phase 3 trial for oncology drug development.

In a recent article, Zhang et al. proposed a 2-in-1 adaptive design to seamlessly expand a selected dose, based on efficacy compared to the control arm, from a Phase 2 trial to a Phase 3 trial for oncology drug development. In this article, we communicate a variation of the proposed design which selects a dose to expand based on direct comparison of high dose to low dose when both doses demonstrate promising efficacy compared to the control arm.

A 2-in-1 adaptive design to seamlessly expand a selected dose from a phase 2 trial to a phase 3 trial for oncology drug development.

In oncology, dose-finding studies are largely performed only in Phase I clinical trials and the maximum tolerated dose (MTD), a dose initially developed for systemic chemotherapies, is by default selected for the Phase 3 confirmatory trial. With the advent of anti-cancer therapies such as molecular targeted agents and immunotherapies, a paradigm shift is underway from the use of conventional MTD approaches to improved dose selection strategies for oncology programs. In response to this new challenge, new study designs are needed to optimize dose selection while still bring life-changing new therapies to patients as soon as possible. In this paper, we propose a 2-in-1 adaptive design starting with a Phase 2 trial with randomized evaluation of multiple doses and only select one dose to expand to a Phase 3 trial if efficacy evidence is observed based on an interim evaluation. The lowest dose will be selected if multiple doses show promising efficacy unless the higher dose demonstrates a more compelling treatment effect, and study will be seamlessly expanded to a Phase 3 trial with the selected dose with patients enrolled in the Phase 2 portion also used for the statistical inference in the Phase 3 portion. The overall Type I error can be controlled under a mild assumption. Simulation studies are conducted to confirm the control of Type I error and to demonstrate the desirable operating characteristics of the proposed design.

Automated Identification of Skull Fractures With Deep Learning: A Comparison Between Object Detection and Segmentation Approach.

Objective: Skull fractures caused by head trauma can lead to life-threatening complications. Hence, timely and accurate identification of fractures is of great importance. Therefore, this study aims to develop a deep learning system for automated identification of skull fractures from cranial computed tomography (CT) scans. Method: This study retrospectively analyzed CT scans of 4,782 patients (median age, 54 years; 2,583 males, 2,199 females; development set: n = 4,168, test set: n = 614) diagnosed with skull fractures between September 2016 and September 2020. Additional data of 7,856 healthy people were included in the analysis to reduce the probability of false detection. Skull fractures in all the scans were manually labeled by seven experienced neurologists. Two deep learning approaches were developed and tested for the identification of skull fractures. In the first approach, the fracture identification task was treated as an object detected problem, and a YOLOv3 network was trained to identify all the instances of skull fracture. In the second approach, the task was treated as a segmentation problem and a modified attention U-net was trained to segment all the voxels representing skull fracture. The developed models were tested using an external test set of 235 patients (93 with, and 142 without skull fracture). Results: On the test set, the YOLOv3 achieved average fracture detection sensitivity and specificity of 80.64, and 85.92%, respectively. On the same dataset, the modified attention U-Net achieved a fracture detection sensitivity and specificity of 82.80, and 88.73%, respectively. Conclusion: Deep learning methods can identify skull fractures with good sensitivity. The segmentation approach to fracture identification may achieve better results.

Health-Related Quality of Life of Patients with Recurrent or Metastatic Cutaneous Squamous Cell Carcinoma Treated with Pembrolizumab in KEYNOTE-629.

INTRODUCTION: Pembrolizumab provided durable responses and acceptable safety in recurrent or metastatic (R/M) cutaneous squamous cell carcinoma (cSCC) in the KEYNOTE-629 study. In this elderly, fragile population with disfiguring tumours, preservation of health-related quality of life (HRQoL) is critical. Here, we present pre-specified exploratory HRQoL analyses from the first interim analysis of KEYNOTE-629. METHODS: Patients with R/M cSCC not amenable to surgery or radiation therapy received pembrolizumab 200 mg every 3 weeks for ≤ 24 months. HRQoL end points included change from baseline to week 12 in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) global health status (GHS)/QoL, functioning, symptom and European Quality of Life 5-Dimension 5-Level (EQ-5D-5L) scores and change from baseline through week 48 in EORTC QLQ-C30 GHS/QoL and physical functioning scores. Improvement (≥ 10-point increase post-baseline with confirmation) was assessed using the exact binomial method. RESULTS: Analyses included 99 patients for EORTC QLQ-C30 and 100 for EQ-5D-5L. Compliance was > 80% at week 12. Mean scores were stable from baseline to week 12 for GHS/QoL (4.95 points; 95% confidence interval, -1.00 to 10.90) and physical functioning (-3.38 points; 95% confidence interval, -8.80 to 2.04). EORTC-QLQ-C30 functioning, symptom, and EQ-5D-5L scores remained stable at week 12. Post-baseline scores were improved in 29.3% of patients for GHS/QoL, 17.2% for physical functioning, and in a numerically higher proportion of responders versus non-responders (GHS/QoL, 55.6% versus 16.1%; physical functioning, 36.1% versus 7.1%). CONCLUSIONS: In elderly patients with R/M cSCC, the clinical efficacy of pembrolizumab translates into a benefit validated by HRQoL preservation or improvement during treatment. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT03284424.

Cutaneous squamous cell carcinoma (cSCC) is the second most common type of non-melanoma skin cancer. cSCC is usually caused by cumulative exposure to sunlight and often occurs in exposed parts of the body such as the head and neck. cSCC is most often seen in older people. If cSCC is detected early, it can be removed by surgery; however, if left untreated, the cancer can spread throughout the body and cause death. The disease itself and its treatment can be painful, cause scarring, or change the patient's physical appearance. Hence, people with cSCC often have poor quality of life. It is therefore important to develop new drugs to help patients with cSCC live longer without worsening their quality of life. The phase 2 KEYNOTE-629 study investigated how well the drug pembrolizumab treated cSCC and whether it was safe. KEYNOTE-629 included patients who were mostly older and had advanced cSCC. The results showed that pembrolizumab was effective and safe. Here, we investigated how pembrolizumab affected the quality of life of these patients. To do this, we asked patients to answer questionnaires on important aspects of their experience, such as their general health status, physical functioning, emotional wellbeing, and symptoms. We found that patients who were treated with pembrolizumab had stable quality of life during treatment. Furthermore, patients whose cancer responded well to pembrolizumab were more likely to have an improved quality of life. These results support the use of pembrolizumab in patients with advanced cSCC.

Adjuvant Pembrolizumab after Nephrectomy in Renal-Cell Carcinoma.

BACKGROUND: Patients with renal-cell carcinoma who undergo nephrectomy have no options for adjuvant therapy to reduce the risk of recurrence that have high levels of supporting evidence. METHODS: In a double-blind, phase 3 trial, we randomly assigned, in a 1:1 ratio, patients with clear-cell renal-cell carcinoma who were at high risk for recurrence after nephrectomy, with or without metastasectomy, to receive either adjuvant pembrolizumab (at a dose of 200 mg) or placebo intravenously once every 3 weeks for up to 17 cycles (approximately 1 year). The primary end point was disease-free survival according to the investigator's assessment. Overall survival was a key secondary end point. Safety was a secondary end point. RESULTS: A total of 496 patients were randomly assigned to receive pembrolizumab, and 498 to receive placebo. At the prespecified interim analysis, the median time from randomization to the data-cutoff date was 24.1 months. Pembrolizumab therapy was associated with significantly longer disease-free survival than placebo (disease-free survival at 24 months, 77.3% vs. 68.1%; hazard ratio for recurrence or death, 0.68; 95% confidence interval [CI], 0.53 to 0.87; P = 0.002 [two-sided]). The estimated percentage of patients who remained alive at 24 months was 96.6% in the pembrolizumab group and 93.5% in the placebo group (hazard ratio for death, 0.54; 95% CI, 0.30 to 0.96). Grade 3 or higher adverse events of any cause occurred in 32.4% of the patients who received pembrolizumab and in 17.7% of those who received placebo. No deaths related to pembrolizumab therapy occurred. CONCLUSIONS: Pembrolizumab treatment led to a significant improvement in disease-free survival as compared with placebo after surgery among patients with kidney cancer who were at high risk for recurrence. (Funded by Merck Sharp and Dohme, a subsidiary of Merck; KEYNOTE-564 ClinicalTrials.gov number, NCT03142334.).

Artificial Intelligence Can Effectively Predict Early Hematoma Expansion of Intracerebral Hemorrhage Analyzing Noncontrast Computed Tomography Image.

This study aims to develop and validate an artificial intelligence model based on deep learning to predict early hematoma enlargement (HE) in patients with intracerebral hemorrhage. A total of 1,899 noncontrast computed tomography (NCCT) images of cerebral hemorrhage patients were retrospectively analyzed to establish a predicting model and 1,117 to validate the model. And a total of 118 patients with intracerebral hemorrhage were selected based on inclusion and exclusion criteria so as to validate the value of the model for clinical prediction. The baseline noncontrast computed tomography images within 6 h of intracerebral hemorrhage onset and the second noncontrast computed tomography performed at 24 ± 3 h from the onset were used to evaluate the prediction of intracerebral hemorrhage growth. In validation dataset 1, the AUC was 0.778 (95% CI, 0.768-0.786), the sensitivity was 0.818 (95% CI, 0.790-0.843), and the specificity was 0.601 (95% CI, 0.565-0.632). In validation dataset 2, the AUC was 0.780 (95% CI, 0.761-0.798), the sensitivity was 0.732 (95% CI, 0.682-0.788), and the specificity was 0.709 (95% CI, 0.658-0.759). The sensitivity of intracerebral hemorrhage hematoma expansion as predicted by an artificial intelligence imaging system was 89.3%, with a specificity of 77.8%, a positive predictive value of 55.6%, a negative predictive value of 95.9%, and a Yoden index of 0.671, which were much higher than those based on the manually labeled noncontrast computed tomography signs. Compared with the existing prediction methods through computed tomographic angiography (CTA) image features and noncontrast computed tomography image features analysis, the artificial intelligence model has higher specificity and sensitivity in the prediction of early hematoma enlargement in patients with intracerebral hemorrhage.

An adaptive seamless Phase 2-3 design with multiple endpoints.

Adaptive seamless Phase 2-3 design has been considered as one possible way to expedite development time for a drug program by allowing the expansion from an ongoing Phase 2 trial into a Phase 3 trial. Multiple endpoints are often tested when a regulatory approval is pursued. Here we propose an adaptive seamless Phase 2-3 design with multiple endpoints which can expand an ongoing Phase 2 trial into a Phase 3 trial based on an intermediate endpoint for adaptive decision and test the endpoints with a powerful multiple test procedure. It is proved that the proposed design can preserve the familywise Type I error under a mild assumption that is expected to hold in practical considerations. We illustrate our proposed design with an example trial design for oncology. Simulations are conducted to confirm the control of the familywise Type I error and the adaptive seamless Phase 2-3 design is illustrated with an example.

A nonparametric method for value function guided subgroup identification via gradient tree boosting for censored survival data.

In randomized clinical trials with survival outcome, there has been an increasing interest in subgroup identification based on baseline genomic, proteomic markers, or clinical characteristics. Some of the existing methods identify subgroups that benefit substantially from the experimental treatment by directly modeling outcomes or treatment effect. When the goal is to find an optimal treatment for a given patient rather than finding the right patient for a given treatment, methods under the individualized treatment regime framework estimate an individualized treatment rule that would lead to the best expected clinical outcome as measured by a value function. Connecting the concept of value function to subgroup identification, we propose a nonparametric method that searches for subgroup membership scores by maximizing a value function that directly reflects the subgroup-treatment interaction effect based on restricted mean survival time. A gradient tree boosting algorithm is proposed to search for the individual subgroup membership scores. We conduct simulation studies to evaluate the performance of the proposed method and an application to an AIDS clinical trial is performed for illustration.

Pembrolizumab Monotherapy for Recurrent or Metastatic Cutaneous Squamous Cell Carcinoma: A Single-Arm Phase II Trial (KEYNOTE-629).

PURPOSE: Treatment options are limited for patients with recurrent and/or metastatic (R/M) cutaneous squamous cell carcinoma (cSCC); mortality rates exceed 70% in patients with distant metastases. Here, we present the first interim analysis of the R/M cSCC cohort from the 2-cohort-locally advanced and R/M-phase II KEYNOTE-629 study. PATIENTS AND METHODS: Patients with R/M cSCC not amenable to surgery or radiation received pembrolizumab 200 mg every 3 weeks. The primary end point was objective response rate per RECIST v1.1. Secondary end points were duration of response, disease control rate, progression-free survival, overall survival, and safety. RESULTS: At data cutoff (April 8, 2019), median follow-up of 105 enrolled patients in the R/M cohort was 11.4 months (range, 0.4 to 16.3 months). Objective response rate was 34.3% (95% CI, 25.3% to 44.2%; 4 complete responses, 32 partial responses), and disease control rate was 52.4% (95% CI, 42.4% to 62.2%). Median duration of response was not reached (range, 2.7 to 13.1+ months; '+' refers to ongoing response at data cutoff). Median progression-free survival was 6.9 months (95% CI, 3.1 months to 8.5 months). Median overall survival was not reached (95% CI, 10.7 months to not reached). Treatment-related adverse events occurred in 66.7% of patients (n = 70), the most common of which were pruritus (n = 15; 14.3%), asthenia (n = 14; 13.3%), and fatigue (n = 13; 12.4%). Grade 3 to 5 treatment-related adverse events occurred in 5.7% (n = 6) of patients. One patient died of treatment-related cranial nerve neuropathy. CONCLUSION: Pembrolizumab demonstrated effective antitumor activity; clinically meaningful, durable responses; and acceptable safety in primarily elderly patients with R/M cSCC, supporting its use in clinical practice. Pembrolizumab adverse events in this study were consistent with its established safety profile.

Pembrolizumab versus methotrexate, docetaxel, or cetuximab for recurrent or metastatic head-and-neck squamous cell carcinoma (KEYNOTE-040): a randomised, open-label, phase 3 study.

BACKGROUND: There are few effective treatment options for patients with recurrent or metastatic head-and-neck squamous cell carcinoma. Pembrolizumab showed antitumour activity and manageable toxicity in early-phase trials. We aimed to compare the efficacy and safety of pembrolizumab versus standard-of-care therapy for the treatment of head-and-neck squamous cell carcinoma. METHODS: We did a randomised, open-label, phase 3 study at 97 medical centres in 20 countries. Patients with head-and-neck squamous cell carcinoma that progressed during or after platinum-containing treatment for recurrent or metastatic disease (or both), or whose disease recurred or progressed within 3-6 months of previous multimodal therapy containing platinum for locally advanced disease, were randomly assigned (1:1) in blocks of four per stratum with an interactive voice-response and integrated web-response system to receive pembrolizumab 200 mg every 3 weeks intravenously or investigator's choice of standard doses of methotrexate, docetaxel, or cetuximab intravenously (standard-of-care group). The primary endpoint was overall survival in the intention-to-treat population. Safety was analysed in the as-treated population. This trial is registered with ClinicalTrials.gov, number NCT02252042, and is no longer enrolling patients. FINDINGS: Between Dec 24, 2014, and May 13, 2016, 247 patients were randomly allocated to pembrolizumab and 248 were randomly allocated to standard of care. As of May 15, 2017, 181 (73%) of 247 patients in the pembrolizumab group and 207 (83%) of 248 patients in the standard-of-care group had died. Median overall survival in the intention-to-treat population was 8·4 months (95% CI 6·4-9·4) with pembrolizumab and 6·9 months (5·9-8·0) with standard of care (hazard ratio 0·80, 0·65-0·98; nominal p=0·0161). Fewer patients treated with pembrolizumab than with standard of care had grade 3 or worse treatment-related adverse events (33 [13%] of 246 vs 85 [36%] of 234). The most common treatment-related adverse event was hypothyroidism with pembrolizumab (in 33 [13%] patients) and fatigue with standard of care (in 43 [18%]). Treatment-related death occurred in four patients treated with pembrolizumab (unspecified cause, large intestine perforation, malignant neoplasm progression, and Stevens-Johnson syndrome) and two patients treated with standard of care (malignant neoplasm progression and pneumonia). INTERPRETATION: The clinically meaningful prolongation of overall survival and favourable safety profile of pembrolizumab in patients with recurrent or metastatic head and neck squamous cell carcinoma support the further evaluation of pembrolizumab as a monotherapy and as part of combination therapy in earlier stages of disease. FUNDING: Merck Sharp & Dohme, a subsidiary of Merck & Co.

Detecting Gene-Environment Interactions for a Quantitative Trait in a Genome-Wide Association Study.

A genome-wide association study (GWAS) typically is focused on detecting marginal genetic effects. However, many complex traits are likely to be the result of the interplay of genes and environmental factors. These SNPs may have a weak marginal effect and thus unlikely to be detected from a scan of marginal effects, but may be detectable in a gene-environment (G × E) interaction analysis. However, a genome-wide interaction scan (GWIS) using a standard test of G × E interaction is known to have low power, particularly when one corrects for testing multiple SNPs. Two 2-step methods for GWIS have been previously proposed, aimed at improving efficiency by prioritizing SNPs most likely to be involved in a G × E interaction using a screening step. For a quantitative trait, these include a method that screens on marginal effects [Kooperberg and Leblanc, 2008] and a method that screens on variance heterogeneity by genotype [Paré et al., 2010] In this paper, we show that the Paré et al. approach has an inflated false-positive rate in the presence of an environmental marginal effect, and we propose an alternative that remains valid. We also propose a novel 2-step approach that combines the two screening approaches, and provide simulations demonstrating that the new method can outperform other GWIS approaches. Application of this method to a G × Hispanic-ethnicity scan for childhood lung function reveals a SNP near the MARCO locus that was not identified by previous marginal-effect scans.

Finding novel genes by testing G × E interactions in a genome-wide association study.

In a genome-wide association study (GWAS), investigators typically focus their primary analysis on the direct (marginal) associations of each single nucleotide polymorphism (SNP) with the trait. Some SNPs that are truly associated with the trait may not be identified in this scan if they have a weak marginal effect and thus low power to be detected. However, these SNPs may be quite important in subgroups of the population defined by an environmental or personal factor, and may be detectable if such a factor is carefully considered in a gene-environment (G × E) interaction analysis. We address the question "Using a genome wide interaction scan (GWIS), can we find new genes that were not found in the primary GWAS scan?" We review commonly used approaches for conducting a GWIS in case-control studies, and propose a new two-step screening and testing method (EDG×E) that is optimized to find genes with a weak marginal effect. We simulate several scenarios in which our two-step method provides 70-80% power to detect a disease locus while a marginal scan provides less than 5% power. We also provide simulations demonstrating that the EDG×E method outperforms other GWIS approaches (including case only and previously proposed two-step methods) for finding genes with a weak marginal effect. Application of this method to a G × Sex scan for childhood asthma reveals two potentially interesting SNPs that were not identified in the marginal-association scan. We distribute a new software program (G×Escan, available at http://biostats.usc.edu/software) that implements this new method as well as several other GWIS approaches.