A comparison of the prognostic performance of the Lugano 2014 and RECIL 2017 response criteria in patients with NHL from the phase III GOYA and GALLIUM trials

The Lugano 2014 criteria are the standard for response assessment in lymphoma. We compared the prognostic performance of Lugano 2014 and the more recently developed response evaluation criteria in lymphoma (RECIL 2017), which relies primarily on computed tomography and uses unidimensional measurements, in patients with previously untreated diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma (FL) from the phase III GOYA and GALLIUM trials, respectively. Concordance between responses according to the Lugano 2014 and RECIL 2017 criteria was analyzed. Landmark analyses of progression-free survival (PFS) and overall survival (OS) by end of treatment (EOT) and end of induction (EOI) response status according to RECIL 2017 and Lugano 2014 criteria, and prognostic value of response at EOT/EOI were also compared. Overall, 1333 patients were included from GOYA and 502 from GALLIUM. Complete response (CR) status according to RECIL 2017 criteria showed high concordance with complete metabolic response (CMR) status by Lugano 2014 criteria in both GOYA (92.5%) and GALLIUM (92.4%). EOT and EOI CR/CMR status by both criteria was highly prognostic for PFS in GOYA (RECIL 2017 [CR]: hazard ratio [HR], 0.35 [95% confidence interval [CI] 0.26-0.46]; Lugano 2014 [CMR]: HR, 0.35 [95% CI 0.26-0.48]; both p < .0001) and GALLIUM (RECIL 2017 [CR]: HR, 0.35 [95% CI 0.23-0.53]; Lugano 2014 [CMR]: HR, 0.21 [95% CI 0.14-0.31]; both p < .0001). In conclusion, response categorization by RECIL 2017 is similar to that by Lugano 2014 criteria, with high concordance observed. Both were prognostic for PFS and OS.

PET-CR as a potential surrogate endpoint in untreated DLBCL: meta-analysis and implications for clinical trial design

This study's focus is the association of end-of-therapy (EOT) PET results with progression-free (PFS) and overall survival (OS) in patients with diffuse large B-cell lymphoma receiving first-line chemoimmunotherapy. We develop a Bayesian hierarchical model for predicting PFS and OS from EOT PET-complete response (PET-CR) using a literature-based meta-analysis of 20 treatment arms and a substudy of 4 treatment arms in 3 clinical trials for which we have patient-level data. The PET-CR rate in our substudy was 72%. The modeled estimates for hazard ratio (PET-CR/non-PET-CR) were 0.13 for PFS (95% CI 0.10, 0.16) and 0.10 for OS (CI 0.07, 0.12). Hazard ratios varied little by patient subtype and were confirmed by the overall meta-analysis. We link these findings to designing future clinical trials and show how our model can be used in adapting the sample size of a trial to accumulating results regarding treatment benefits on PET-CR and a survival endpoint.

A prognostic model integrating PET‐derived metrics and image texture analyses with clinical risk factors from GOYA

Image texture analysis (radiomics) uses radiographic images to quantify characteristics that may identify tumour heterogeneity and associated patient outcomes. Using fluoro‐deoxy‐glucose positron emission tomography/computed tomography (FDG‐PET/CT)‐derived data, including quantitative metrics, image texture analysis and other clinical risk factors, we aimed to develop a prognostic model that predicts survival in patients with previously untreated diffuse large B‐cell lymphoma (DLBCL) from GOYA (NCT01287741). Image texture features and clinical risk factors were combined into a random forest model and compared with the international prognostic index (IPI) for DLBCL based on progression‐free survival (PFS) and overall survival (OS) predictions. Baseline FDG‐PET scans were available for 1263 patients, 832 patients of these were cell‐of‐origin (COO)‐evaluable. Patients were stratified by IPI or radiomics features plus clinical risk factors into low‐, intermediate‐ and high‐risk groups. The random forest model with COO subgroups identified a clearer high‐risk population (45% 2‐year PFS [95% confidence interval (CI) 40%–52%]; 65% 2‐year OS [95% CI 59%–71%]) than the IPI (58% 2‐year PFS [95% CI 50%–67%]; 69% 2‐year OS [95% CI 62%–77%]). This study confirms that standard clinical risk factors can be combined with PET‐derived image texture features to provide an improved prognostic model predicting survival in untreated DLBCL.

Total metabolic tumor volume as a survival predictor for patients with diffuse large B-cell lymphoma in the GOYA study

This retrospective analysis of the phase III GOYA study investigated the prognostic value of baseline metabolic tumor volume parameters and maximum standardized uptake values for overall and progression-free survival (PFS) in treatment-naïve diffuse large B-cell lymphoma. Baseline total metabolic tumor volume (determined for tumors >1 mL using a threshold of 1.5 times the mean liver standardized uptake value +2 standard deviations), total lesion glycolysis, and maximum standardized uptake value positron emission tomography data were dichotomized based on receiver operating characteristic analysis and divided into quartiles by baseline population distribution. Of 1,418 enrolled patients, 1,305 had a baseline positron emission tomography scan with detectable lesions. Optimal cut-offs were 366 cm³ for total metabolic tumor volume and 3,004 g for total lesion glycolysis. High total metabolic tumor volume and total lesion glycolysis predicted poorer PFS, with associations retained after adjustment for baseline and disease characteristics (high total metabolic tumor volume hazard ratio: 1.71, 95% confidence interval [CI]: 1.35– 2.18; total lesion glycolysis hazard ratio: 1.46; 95% CI: 1.15–1.86). Total metabolic tumor volume was prognostic for PFS in subgroups with International Prognostic Index scores 0–2 and 3–5, and those with different cell-of-origin subtypes. Maximum standardized uptake value had no prognostic value in this setting. High total metabolic tumor volume associated with high International Prognostic Index or non-germinal center B-cell classification identified the highest-risk cohort for unfavorable prognosis. In conclusion, baseline total metabolic tumor volume and total lesion glycolysis are independent predictors of PFS in patients with diffuse large B-cell lymphoma after first-line immunochemotherapy.

Follicular Lymphoma Evaluation Index (FLEX): A new clinical prognostic model that is superior to existing risk scores for predicting progression-free survival and early treatment failure after frontline immunochemotherapy

Patients with advanced‐stage follicular lymphoma (FL) who progress early after receiving first‐line therapy have poor overall survival (OS). Currently applied clinical prognostic models such as FL International Prognostic Index [FLIPI], FLIPI‐2 and PRIMA‐Prognostic Index [PRIMA‐PI] have suboptimal sensitivity and specificity to predict this poor prognosis subgroup. The primary objective was to develop a novel prognostic model, the FL Evaluation Index (FLEX) score, to identify high‐risk patients and compare its performance with FLIPI, FLIPI‐2 and PRIMA‐PI. Progression‐free survival (PFS) after first‐line immunochemotherapy was the key endpoint, while OS and progression of disease within 24 months (POD24) were also assessed. The model, which includes nine clinical variables, was developed using a cohort of patients with previously untreated advanced‐stage FL from the phase 3 GALLIUM trial (NCT01332968). The performance of the model was validated using data from the SABRINA trial (NCT01200758). In GALLIUM (n = 1004; 127 with and 877 without POD24), FLEX increased the intergroup (low‐risk/high‐risk) difference in 2‐year and 3‐year PFS rates and demonstrated superior intergroup differences in 2‐year and 3‐year OS rates compared with FLIPI, FLIPI‐2 and PRIMA‐PI. Sensitivity for a high‐risk score to predict POD24 was 60% using FLEX compared with 53% for FLIPI and FLIPI‐2, and 69% for PRIMA‐PI, while specificity was 68% for FLEX compared with 58% for FLIPI, 59% for FLIPI‐2 and 48% for PRIMA‐PI. The prognostic value of FLEX in SABRINA was similar to FLIPI. Therefore, FLEX appears to perform better than existing prognostic models in previously untreated FL, in particular for the newer treatment regimens.

A broken promise: microbiome differential abundance methods do not control the false discovery rate

High-throughput sequencing technologies allow easy characterization of the human microbiome, but the statistical methods to analyze microbiome data are still in their infancy. Differential abundance methods aim at detecting associations between the abundances of bacterial species and subject grouping factors. The results of such methods are important to identify the microbiome as a prognostic or diagnostic biomarker or to demonstrate efficacy of prodrug or antibiotic drugs. Because of a lack of benchmarking studies in the microbiome field, no consensus exists on the performance of the statistical methods. We have compared a large number of popular methods through extensive parametric and nonparametric simulation as well as real data shuffling algorithms. The results are consistent over the different approaches and all point to an alarming excess of false discoveries. This raises great doubts about the reliability of discoveries in past studies and imperils reproducibility of microbiome experiments. To further improve method benchmarking, we introduce a new simulation tool that allows to generate correlated count data following any univariate count distribution; the correlation structure may be inferred from real data. Most simulation studies discard the correlation between species, but our results indicate that this correlation can negatively affect the performance of statistical methods.

A core laboratory approach to large-scale radiomics and machine-learning prediction of DLBCL outcomes after first-line treatment using results from the phase III GOYA study

e19042

Background: We are using the phase 3 GOYA study to retrospectively investigate using quantitative image texture features (i.e. radiomics, evaluating tumor heterogeneity) to improve the prognostic value of baseline FDG-PET. We report on the considerations for analysis, which is ongoing, and solutions to methodological challenges. Methods: GOYA, an open-label, randomized, phase 3 study, enrolled 1418 patients from 207 centers; 1334 had a baseline FDG-PET scan with detectable lesions. Image data with annotated regions of interest (ROIs) defined by qualified physicians were transferred to the Cuneo imaging core laboratory. Images were interpolated to a common 4x4x4 mm voxel size and ROI masks using 4 different definition methodologies were used to define image sub-volumes that contained the lesions. Large ROIs in normal liver were also defined. The 42 Haralick radiomic image texture features were computed using the PET oncology radiomics test suite. Results: A total of 9286 lesions were analyzed. Interpolation of images to a fixed size is challenging in this cohort due the large number of features to be computed (~1.5M). As a test, the lesions were differentiated from liver ROIs using 5 classifiers (e.g. support vector machine, random forest, etc) with an accuracy of 0.995–0.999. The mean (±SD) of the metabolically active tumor volume was 21.2±22.2 mL (range 0.2–146 mL) with a preponderance of smaller lesions as shown by the median of 15.0 mL. The skewed distribution had downstream effects on many of the quantitative (e.g. total lesion glycolysis) and radiomic image texture features. The majority of texture features were computable; 5 (Coarseness, Contrast, Busyness, Complexity, Texture Strength) required interpolation to a 2x2x2 mm voxel size, requiring extended computation time. FDG-PET images came from a variety of PET/CT scanners; thus, a unified image reconstruction protocol was not possible. Conclusions: Radiomics analysis of large phase 3 imaging trials is feasible. However, careful attention should be paid to the entire processing chain, including tests for non-computable features values due to highly skewed distributions and lesion size. Clinical trial information: NCT01287741.

Overall Survival Benefit in Patients With Rituximab-Refractory Indolent Non-Hodgkin Lymphoma Who Received Obinutuzumab Plus Bendamustine Induction and Obinutuzumab Maintenance in the GADOLIN Study

Purpose To perform an updated analysis of the randomized phase III GADOLIN trial in patients with rituximab-refractory indolent non-Hodgkin lymphoma treated with obinutuzumab (GA101; G) and bendamustine (B). Patients and Methods Patients with histologically documented, rituximab-refractory CD20⁺ indolent non-Hodgkin lymphoma received G 1,000 mg (days 1, 8, and 15, cycle 1; day 1, cycles 2 to 6) plus B 90 mg/m²/d (days 1 and 2, all cycles) or B 120 mg/m²/d monotherapy. Patients who did not experience disease progression with G-B received G maintenance (1,000 mg every 2 months) for up to 2 years. The primary end point was progression-free survival (PFS). Results Of 413 randomly assigned patients (intention-to-treat [ITT]: G-B, n = 204; B monotherapy, n = 209), 335 had follicular lymphoma (FL; G-B, n = 164; B monotherapy, n = 171). After a median follow-up of 31.8 months, median PFS in ITT patients was 25.8 months (G-B) and 14.1 months (B monotherapy; hazard ratio [HR], 0.57; 95% CI, 0.44 to 0.73; P < .001). Overall survival (OS) also was prolonged (HR, 0.67; 95% CI, 0.47 to 0.96; P = .027). PFS and OS benefits were similar in patients with FL. Grade 3 to 5 adverse events (AEs) were reported by 148 (72.5%) and 133 (65.5%) patients in the G-B and B monotherapy arms, respectively, most commonly neutropenia (G-B, 34.8%; B monotherapy, 27.1%), thrombocytopenia (10.8% and 15.8%), anemia (7.4% and 10.8%), and infusion-related reactions (9.3% and 3.4%). Serious AEs occurred in 89 G-B patients (43.6%) and 75 B monotherapy patients (36.9%); fatal AEs occurred in 16 (7.8%) and 13 (6.4%), respectively. Conclusion This updated analysis confirms the PFS benefit for G-B shown in the primary analysis. A substantial OS benefit also was demonstrated in the ITT population and in patients with FL. Toxicity was similar for both treatments.

A web application for sample size and power calculation in case-control microbiome studies

: When designing a case-control study to investigate differences in microbial composition, it is fundamental to assess the sample sizes needed to detect an hypothesized difference with sufficient statistical power. Our application includes power calculation for (i) a recoded version of the two-sample generalized Wald test of the 'HMP' R-package for comparing community composition, and (ii) the Wilcoxon-Mann-Whitney test for comparing operational taxonomic unit-specific abundances between two samples (optional). The simulation-based power calculations make use of the Dirichlet-Multinomial model to describe and generate abundances. The web interface allows for easy specification of sample and effect sizes. As an illustration of our application, we compared the statistical power of the two tests, with and without stratification of samples. We observed that statistical power increases considerably when stratification is employed, meaning that less samples are needed to detect the same effect size with the same power.

AVAILABILITY AND IMPLEMENTATION

The web interface is written in R code using Shiny (RStudio Inc., 2016) and it is available at https://fedematt.shinyapps.io/shinyMB The R code for the recoded generalized Wald test can be found at https://github.com/mafed/msWaldHMP CONTACT: Federico.Mattiello@UGent.be.

Principal bicorrelation analysis: Unraveling associations between three data sources

In this article, we propose a statistical explorative method for data integration. It is developed in the context of early drug development for which it enables the detection of chemical substructures and the identification of genes that mediate their association with the bioactivity (BA). The core of the method is a sparse singular value decomposition for the identification of the gene set and a permutation-based method for the control of the false discovery rate. The method is illustrated using a real dataset, and its properties are empirically evaluated by means of a simulation study. Quantitative Structure Transcriptional Activity Relationship (QSTAR, www.qstar-consortium.org ) is a new paradigm in early drug development that extends QSAR by not only considering data on the chemical structure of the compounds and on the compound-induced BA, but by simultaneously using transcriptomics data (gene expression). This approach enables, for example, the detection of chemical substructures that are associated with BA, while at the same time a gene set is correlated with both these substructures and the BA. Although causal associations cannot be formally concluded, these associations may suggest that the compounds act on the BA through a particular genomic pathway.