Step-down Procedure Research Articles

On limiting behaviors of stepwise multiple testing procedures

AbstractStepwise multiple testing procedures have attracted several statisticians for decades and are also quite popular with statistics users because of their technical simplicity. The Bonferroni procedure has been one of the earliest and most prominent testing rules for controlling the familywise error rate (FWER). A recent article established that the FWER for the Bonferroni method asymptotically (i.e., when the number of hypotheses becomes arbitrarily large) approaches zero under any positively equicorrelated multivariate normal framework. However, similar results for the limiting behaviors of FWER of general stepwise procedures are nonexistent. The present work addresses this gap in a unified manner by elucidating that, under the multivariate normal setups with some correlation structures, the probability of rejecting one or more null hypotheses approaches zero asymptotically for any step-down procedure. Consequently, the FWER and power of the step-down procedures also tend to be asymptotically zero. We also establish similar limiting zero results on FWER of other popular multiple testing rules, e.g., Hochberg’s and Hommel’s procedures. It turns out that, within our chosen asymptotic framework, the Benjamini–Hochberg method can hold the FWER at a strictly positive level asymptotically under the equicorrelated normality.

Multiple exponentially weighted moving average control schemes for monitoring and diagnostics of correlated quality variables of different types with individual observations

In the era of big data analytics, more data, in terms of quantity and variety, are being collected for quality control purposes, thus leading to new challenges, especially for multivariate statistical process control (MSPC). One such challenge, which has received very little attention, is multivariate control chart (MCC) for monitoring correlated quality variables of different types. A recent study by Huang et al. (2023), which develops a Shewhart type MCC, is the first attempt to tackle such a challenge. The proposed chart of Huang et al. (2023), which utilizes the step-down multiple testing procedure of Holm (1979), not only can monitor correlated variables of different types, but also can provide instantaneous diagnostics of which parameters are out of control when the chart signals. In this study, we adapt and extend the methodology of Huang et al. (2023) to develop an exponentially weighted moving average (EWMA) chart specifically for the case when the sample size is one. The proposed chart is shown to be effective in detecting parameter changes as well as diagnosing which parameters are out of control when the chart signals.

The HEARTRISK6 Scale: Predicting Short-Term Serious Outcomes in Emergency Department Acute Heart Failure Patients

BackgroundAcute heart failure (AHF) is a common emergency department (ED) presentation that may have poor outcomes but often does not require hospital admission. There is little evidence to guide dispositional decisions. ObjectivesThe authors sought to create a risk score for predicting short-term serious outcomes (SSO) in patients with AHF. MethodsWe pooled data from 3 prospective cohorts: 2 published studies and 1 new cohort. The 3 cohorts prospectively enrolled patients who required treatment for AHF at 10 tertiary care hospital EDs. The primary outcome was SSO, defined as death <30 days, intubation or noninvasive ventilation (NIV), myocardial infarction, or relapse to ED <14 days. The logistic regression model evaluated 13 predictors, used an AIC-based step-down procedure, and bootstrapped internal validation. ResultsOf the 2,246 patients in the 3 cohorts (N = 559; 1,100; 587), the mean age was 77.4 years, 54.5% were male, 3.1% received intravenous nitroglycerin, 5.2% received ED NIV, and 48.6% were admitted to the hospital. There were 281 (12.5%) SSOs including 70 deaths (3.1%) with many in discharged patients. The final HEARTRISK6 Scale included 6 variables: valvular heart disease, tachycardia, need for NIV, creatinine, troponin, and failed reassessment (walk test). Choosing HEARTRISK6 total-point admission thresholds of ≥1 or ≥2 would yield, respectively, sensitivities of 88.3% (95% CI: 83.9%-91.8%) and 71.5% (95% CI: 65.9%-76.7%) and specificities of 24.7% (95% CI: 22.8%-26.7%) and 50.1% (95% CI: 47.9%-52.4%) for SSO. ConclusionsUsing 3 large prospectively collected datasets, we created a concise and sensitive risk scale for patients with AHF in the ED. Implementation of the HEARTRISK6 scale could lead to safer and more efficient disposition decisions.

12 Time to Sustained Recovery between Oral Tablet and Inhaler Placebos in the ACTIV-6 Platform Clinical Trial

OBJECTIVES/GOALS: Platform trials gain efficiency by sharing placebo controls among different study arms. However, the varying routes of administration make it unclear whether participants exposed to different placebos have similar outcomes. As such, we seek to compare outcomes between participants receiving tablet and inhaler placebos in the ACTIV-6 trial. METHODS/STUDY POPULATION: ACTIV-6 is a large, decentralized platform trial exploring repurposed drugs for the treatment of adults with mild to moderate COVID-19. Enrolled participants were randomly assigned to a study arm vs. placebo and then mailed the study drug. They were monitored until symptom resolution or Day 28. Here, we compare outcomes for control participants contributing to the fluticasone furoate study arm, in which 251 were assigned to a tablet placebo and 370 an inhaler placebo. Time to sustained recovery and time to resolution of individual symptoms are compared between groups using Kaplan-Meier curves and unadjusted log-rank tests. A step-down procedure is applied to control the false discovery rate. RESULTS/ANTICIPATED RESULTS: Control participants assigned to tablet placebos had shorter time to sustained recovery (adjusted hazard ratio (HR) 1.34 (95% CI 1.11, 1.62)). When examining each of the eleven individually reported symptoms on study Day 14, nasal symptoms (adjusted odds ratio (OR) 0.44 (0.27, 0.72), p&lt;0.01), dyspnea (OR 0.44 (0.22, 0.87), p = 0.02), and cough (OR 0.54 (0.35, 0.83), p&lt;0.01) were identified as symptoms in which the tablet-placebo group performed notably better than those who received inhaler-placebos. In the follow-up, longitudinal analysis, we anticipate similar results. DISCUSSION/SIGNIFICANCE: Among ACTIV-6 control participants, those receiving a tablet placebo had a significantly shorter time to sustained recovery than those receiving an inhaler placebo. Platform trials using shared controls should consider efficiency in the context of the additional variability when sharing controls with a different route of administration.

The Association Between Blood Mercury and Lipid Biomarkers in US Hypertensive Adults.

This cross-sectional research gathered data from a total of 4415 participants sourced from the National Health and Nutrition Examination Survey (NHANES). The Holm-Bonferroni stepdown procedure was utilized to control the type I error rate in multiple comparisons. We employed multivariable linear regression models to assess the correlation between blood Hg and lipid biomarkers. Subsequently, subgroup analyses were conducted, categorized by both gender and race. Additionally, we used smooth curve fittings and generalized additive models to confirm the presence of non-linear relationships. When non-linearity was detected, we applied a recursive algorithm to calculate the inflection points. Finally, we established a weighted two-piecewise linear regression model to illustrate the associations on either side of the inflection point. In our multivariable linear regression models, clear associations emerged. Specifically, positive correlations were observed between blood mercury and TC (β = 0.025; 95% CI 0.009 to 0.041; corrected P = 0.011), LDL-C (β = 0.022; 95% CI 0.007 to 0.036; corrected P = 0.012), and HDL-C (β = 0.007; 95% CI 0.001 to 0.013; corrected P = 0.058). However, there was no significant correlation with TG (β = - 0.007; 95% CI - 0.018 to 0.004; corrected P = 0.526). Notably, it has been demonstrated that distinct inverted U-shaped and U-shaped curves exist when stratified by gender in our analysis. Blood Hg exhibited a positive correlation with TC, LDL-C, and HDL-C in hypertensive adults in the USA. Nonetheless, no significant association was observed with TG.

Financial Toxicity in the Spanish Speaking Population

Advances in therapy have significantly improved cancer survival outcomes but with a harmful patient consequence: financial toxicity. Financial toxicity consists of a patient's many economic burdens from cancer treatment, encompassing direct healthcare costs as well as additional devastating financial consequences, such as job loss and medical bankruptcy. With rising cancer treatment costs, financial toxicity has become a problem that is increasingly impactful in both cancer treatment and survivorship, amplifying disparities, negatively impacting quality of life, treatment adherence and survival. The long-term research goal that this study will contribute to is to develop culturally appropriate and effective patient navigation resources and psychosocial support resources for interventions to increase coping and resilience to financial toxicity in high-risk populations. This study will seek to generate the early data needed to help develop the content and method of delivering strategies, tailored for an underserved, Spanish-speaking Latino population. Eligibility Participant must be 18 years of age or older at the time of study participation, able to read and speak Spanish, have diagnosis of cancer within the past 2 years or be on active treatment for their cancer; all disease sites and all stages are eligible for enrollment. We will exclude patients unable to give consent. Methods This is a prospective observational study of cancer patients and survivors, with participants to be recruited from Scripps MD Anderson Cancer Center. Participating clinics will be screened in the electronic medical record for individuals who meet above eligibility criteria. At the time of clinic visits, eligible individuals will be approached. Individuals who express that they would like to participate in the survey study will receive the study packet with the consent statement and baseline survey questions. Study participants will also be invited to participate in the OPTIONAL patient navigation portion of the study. Participants who are not active on the MyScripps patient portal will be invited to join the patient navigation portion of the study. Those elect to participate will be provided a device if they do not have their own. Participants will be given a Greenphire Clincard on which they will receive a total of $75 in three installments. Research team members will enter the item responses into the REDCap database. They will undergo a follow-up survey during a range from 4 months to 9 months from the initial survey date. Those participating in the patient navigation portion of the study will be contacted by the navigator on a monthly basis through the electronic portal or telephone. Statistics We plan to enroll N= 80 Spanish-speaking patients in the survey study. For correlation testing, we hypothesize a moderate effect size (correlation = 0.32) for the association between the primary outcome measure, the baseline measure of self-efficacy (Hope State score) and the baseline financial toxicity score (Economic Strain and Resilience in Cancer). For feasibility testing of navigation participation, feasibility will be assessed through descriptive statistics. We hypothesize that it is feasible for 50% of invited patients to participate. When the sample size is 40, a two-sided 95% confidence interval for participant completion rate will extend 0.15 from the observed proportion for an expected proportion of 0.50. Analysis The primary outcome measure is financial toxicity, measured using the ENRICh instrument at baseline. The primary predictor is the baseline Hope State score. This modeled association is of primary interest in the analysis. We will conduct descriptive and univariate/multivariate analyses on this association of primary interest, including Chi-square test, t-test. Other univariate and multivariate analyses of associations between financial toxicity and other measures obtained from survey measures, as well as clinical, demographic, socioeconomic, and care delivery variables, considered descriptive and exploratory, will be performed. Accounting for multiple comparisons will be based on stepdown procedures and familywise error rate control; we plan to control this overall error rate using the Holm-Sidak adjustment, or procedure with equitable properties.

GEM2017FIT Trial: Induction Therapy with Bortezomib-Melphalan and Prednisone (VMP) Followed By Lenalidomide and Dexamethasone (Rd) Versus Carfilzomib, Lenalidomide and Dexamethasone (KRd) Plus/Minus Daratumumab (D), 18 Cycles, Followed By Consolidation and Maintenance Therapy with Lenalidomide and Daratumumab: Phase III, Multicenter, Randomized Trial for Elderly Fit Newly Diagnosed Multiple Myeloma (NDMM) Patients Aged between 65 and 80 Years

Background: In transplant-ineligible patients (pts) with NDMM, VMP and Rd have been standards of care. The Spanish Myeloma Group examined induction with VMP followed by Rd, and results were particularly striking for “fit” pts aged 65-80 with a CR rate of 47% and an MRD negative rate (MRD(-)) of 20%. The median PFS was 34 months (m) and 4-years (yrs) OS rate was 70%. Pts and study design: Based on this background, we consider VMP9-Rd9 as control arm and compared with KRd or D-KRd. Induction included 18 cycles, followed by 4 D-Rd consolidation cycles for the two first arms (to evaluate if anti CD38 consolidation might compensate its absence as induction). Subsequently, pts were stratified by MRD (positive and negative) to maintenance therapy with D-R vs observation. Frailty was evaluated with Geriatric Assessment in Hematology (GAH scale) considering as fit those younger than 80 years and with a punctuation between 0 and 42. VMP and Rd were given at conventional doses and in the K-based regimens, K was given at dose of 36mg/m 2 twice weekly during C1-2 and at 56 mg/m 2 weekly since C3. R was given at standard dose and D was planned as IV but switched to SC, both at the conventional dose and schedule. The primary endpoint was to evaluate the MRD(-) rate (NGF 10 -5) at the end of induction to compare both experimental arms, KRd and D-KRd to the control-one, VMP-Rd. A step-down testing procedure was applied, whereby inference for a test in the pre-defined hierarchy was dependent upon statistical significance having been achieved for the previous tests in the hierarchy. No formal adjustment for multiplicity of contrasts is required. Secondary endpoints referred to the induction phase included the serological responses, PFS and OS probabilities at 18 cycles and safety profile. Results: 540 pts were screened and 78 were screening failures. 462 pts were randomized to VMP-Rd (154 pts), KRd (154 pts) and D-KRd (153 pts but one patient was not evaluable). Baseline characteristics of the pts were comparable in the three arms with a median age of 72 and around one third older than 75 years. Nearly one third of pts in the three arms had ISS 3 and 15% had extramedullary disease. A total of 369 out of 461 pts have completed the 18 induction cycles (128 in VMP-Rd, 122 in KRd and 119 in D-KRd) and were evaluable for the primary endpoint. The MRD(-) rate at 10 -5 was 32% for VMP-Rd, 69% for KRd (p&amp;lt;0.0001) and 79% for D-KRd (p&amp;lt;0.0001). At 10 -6, the MRD(-) rate was also significantly superior for KRd (59%) and D-KRd (75%) in comparison with VMP-Rd (24%) (p values &amp;lt;0.0001). The ORR at the end of induction was comparable in the three arms (87%, 89% and 89%, respectively) but the sCR/CR rate was significantly higher for KRd (59%) and D-KRd (61%) in comparison with VMP-Rd (40%) (p-values &amp;lt;0.0001). At 18 m, the proportion of pts alive and progression-free was 79% for VMP-Rd and 87% for KRd and D-KRd, respectively. Only two pts who achieved MRD negativity have progressed thus far. The OS at 18 m in the VMP-Rd, KRd and D-KRd arms was 91%, 95% and 90%, respectively. There were 48, 44 and 40 pts who early discontinued therapy in VMP-Rd, KRd and D-KRd, respectively, due to progressive disease (VMP-Rd: 23; KRd: 9;D-KRd: 6) or toxicity (VMP/Rd:8; KRd: 14; D-KRd: 6). Death occurred in 12 pts treated with D-KRd being infections the cause in 10 pts (Covid-19 in 3); 5 pts in the KRd arm (4 infections (Covid-19 in 1)) and 6 pts in VMP-Rd arm (infections in 4 pts). Concerning hematological toxicity during induction, G3/4 neutropenia was higher in VMP-Rd (50%) than KRd (24%) (p&amp;lt;0.0001) and similar to D-KRd (47%). G3/4 thrombocytopenia was also higher in VMP-Rd (34%) than the other two arms (16% and 17%, p&amp;lt;0.0001)). G3/4 infections rate was comparable in the three arms (12% in VMP-Rd and 15%-16% in KRd and D-KRd). Cardiovascular G3/4 events were higher in KRd (11%) and D-KRd (14%) than in VMP-Rd (5%) (p&amp;lt;0.0001) being hypertension and cardiac failure the most frequent-ones. Conclusions: The trial met its primary end point and the MRD(-) rates of 69% for KRd and 79% for D-KRd were significantly higher than VMP9-Rd9 (32%). Overall, the VMP-Rd arm had higher rates of neutropenia, while the D-KRd arm had higher rates of infections leading to death. The trial is ongoing to evaluate long term outcomes.

Step-down multiple testing procedures for the overall population and partitioning subgroups

ABSTRACT Multiple comparisons procedures are described for two-armed studies where there is a primary hypothesis and one or more ordered secondary hypotheses in the scenario where the objective is to test for an effect on the overall population and/or nonoverlapping subgroups that partition the population. This scenario arises when the subgroups are defined by disease etiology or by other patient characteristics such as genetic factors, age, sex, or race, and the treatment may have differential effects in the subgroups. The procedures described control the family-wise error rate in the strong sense at a specified level α.

Rucaparib for metastatic castration-resistant prostate cancer (mCRPC): TRITON3 interim overall survival and efficacy of rucaparib vs docetaxel or second-generation androgen pathway inhibitor therapy.

18 Background: TRITON3 (NCT02975934) is a randomized, multicenter, open-label, phase 3 study of rucaparib vs physician’s choice (docetaxel [DTX], abiraterone [ABI] or enzalutamide [ENZ]) in patients (pts) with chemotherapy-naïve mCRPC with BRCA1/2 (BRCA) or ATM alterations. It was previously reported that rucaparib significantly improved the primary endpoint of radiographic progression-free survival (rPFS) vs physician’s choice in pts with BRCA or ATM alterations. We report here the interim overall survival (OS) and also report on efficacy of rucaparib compared individually with either DTX or ABI/ENZ. Methods: Pts had disease progression after 1 prior second-generation androgen pathway inhibitor therapy in any setting and were randomized 2:1 to rucaparib 600 mg BID or physician’s choice of DTX, ABI or ENZ. An ordered step-down multiple comparisons procedure was used to control the overall error rate. OS was a key secondary endpoint with rPFS as the primary endpoint tested first in the BRCA subgroup, then the intent-to-treat (ITT) population. Subgroup analyses based on physician’s choice were exploratory. Results: As of August 25, 2022, 302 pts with BRCA and 103 pts with ATM alterations were randomized. OS maturity was 54% in the BRCA subgroup and 59% in the ITT population. rPFS and OS are shown. The most frequent treatment-emergent adverse event (TEAE) in the rucaparib, DTX and ABI/ENZ groups was asthenia/fatigue (61.1%, 67.6% and 57.6%, respectively). The most frequent grade ≥3 TEAE in the rucaparib, DTX and ABI/ENZ groups was anemia (23.7%), neutropenia (14.1%), and hypertension (10.2%), respectively. Conclusions: Rucaparib significantly improved rPFS vs either DTX or ABI/ENZ; safety was consistent with prior reports. Interim OS results suggest a trend towards improvement for rucaparib vs DTX or ABI/ENZ in pts with mCRPC and BRCA alterations. Clinical trial information: NCT02975934 . [Table: see text]

Hemoglobin and Neutrophil Count as Prognostic Factors in Cholangiocarcinoma Patients in 2nd Line Treatment Setting: Results from a Small Monocentric Retrospective Study.

Unresectable cholangiocarcinoma prognosis can be extremely variable due to different symptoms and sites of disease involvement at diagnosis and unpredictable chemotherapy response rates. Most patients will usually receive 1st line palliative chemotherapy with platinum compounds and Gemcitabine or Gemcitabine alone. Only a few patients maintain adequate performance status after first-line treatment failure: second-line treatment with FOLFOX or FOLFIRI chemotherapy has been used in this setting with modest overall survival improvement. There is a lack of data concerning whether laboratory findings might help clinicians in identifying those patients with the highest likelihood of benefiting from 2nd line treatment. The aim of this analysis is to assess the prognostic role of a series of easily available laboratory tests in patients with bile duct cancer who received 2nd line chemotherapy. Patients with unresectable bile duct cancer treated in 2nd-line setting with platinum-based chemotherapy doublet or FOLFIRI were enrolled. The primary objective of the analysis was to assess overall survival (OS) differences among patients based on the results of lab tests. Serum hemoglobin, neutrophil, lymphocyte, monocyte, platelet absolute count, creatinine, total bilirubin, albumin, LDH, circulating CEA and CA19.9 values were collected at the start of 2nd line treatment. Cut-off values for all lab tests were set by ROC curve analysis. Survival was calculated by the Kaplan-Meier method and differences in survival among stratification factors were assessed by Log-rank test. Cox-proportional-hazard regression was used for multivariate analysis. Level of statistical significance p was set at 0.05 for all tests. Correction for false discovery error rate was performed by Holm's stepdown procedure. A total of 46 patients were eligible. Median overall survival of the entire cohort was 8.98 months (95%CI: 6.68-13.93) while mean OS was 17.10 months (standard error: 3.16). Using 6.2 months OS landmark as classification variable for ROC curve analysis, only serum hemoglobin (cut-off: &gt;10 g/dL), albumin (cut-off: &gt;3.5 mg/dL), CA19.9 (cut-off: ≤668 UI/mL), monocyte (cut-off: ≤510/mmc) and neutrophil count (cut-off: ≤5140/mmc) were significantly associated with the chosen end-point. Multivariate analysis confirmed an independent statistically significant impact on overall survival only for hemoglobin (Exp(b): 0.12, p = 0.0023) and neutrophil count (Exp(b): 0.30, p = 0.0039). Based on these results, using both hemoglobin and neutrophil count, three prognostic groups were defined: patients with both favorable factors had 12.63 months median OS vs. 6.75 months of patients with only one favorable factor vs. 1.31 months of those with neither. The difference between these three groups of patients was statistically significant (p &lt; 0.0001). Second-line palliative chemotherapy can be a potentially useful option for a few patients with unresectable/metastatic bile duct cancer. Even though assessment of patients' prognosis might be difficult due to the complex behavior of this disease, a series of easily available laboratory tests might be used for these means: serum hemoglobin and neutrophil count we0re able to define subsets of patients with entirely different prognoses. It is hoped that this score will be prospectively validated in a larger group of patients in order to improve treatment decisions in patients with unresectable bile duct cancer candidate to receive palliative 2nd line chemotherapy.

Monitoring and diagnostics of correlated quality variables of different types

As data acquisition and processing technologies continue to advance rapidly, new challenges emerge for statistical process monitoring. One such challenge, especially in the era of big data analytics, is monitoring multivariate processes involving a mixture of continuous, categorical, and discrete quality variables. The existing multivariate control charts focus mostly on monitoring correlated variables of the same type. We propose a new Phase II control chart that is based on a modified Holm’s step-down multiple testing procedure (Holm 1979) which achieves two important goals at the same time: (1) it simultaneously monitors correlated variables of different types, while keeping the probability of false alarm under desirable level, and (2) when the process is determined to be out of control, it further provides, without any additional efforts, diagnostics to pinpoint which parameters are out of control. The proposed chart is shown to outperform the existing charts particularly in its ability to provide more accurate diagnostics.

Error-Based Knockoffs Inference for Controlled Feature Selection

Recently, the scheme of model-X knockoffs was proposed as a promising solution to address controlled feature selection under high-dimensional finite-sample settings. However, the procedure of model-X knockoffs depends heavily on the coefficient-based feature importance and only concerns the control of false discovery rate (FDR). To further improve its adaptivity and flexibility, in this paper, we propose an error-based knockoff inference method by integrating the knockoff features, the error-based feature importance statistics, and the stepdown procedure together. The proposed inference procedure does not require specifying a regression model and can handle feature selection with theoretical guarantees on controlling false discovery proportion (FDP), FDR, or k-familywise error rate (k-FWER). Empirical evaluations demonstrate the competitive performance of our approach on both simulated and real data.

ATHENA–MONO (GOG-3020/ENGOT-ov45): A randomized, double-blind, phase 3 trial evaluating rucaparib monotherapy versus placebo as maintenance treatment following response to first-line platinum-based chemotherapy in ovarian cancer.

LBA5500 Background: While PARP inhibitors have shown efficacy as first-line (1L) maintenance treatment for patients (pts) with ovarian cancer (OC), questions remain about the pt population that may benefit from their use. ATHENA (NCT03522246) was designed to test if rucaparib may be effective as 1L maintenance treatment in a broad pt population, including those without BRCA mutations or other evidence of homologous recombination deficiency (HRD), or high-risk clinical characteristics such as residual disease. Here we report results from the ATHENA–MONO comparison of rucaparib vs placebo. Methods: Pts with stage III–IV high-grade OC who had completed cytoreductive surgery (R0 permitted) and 4–8 cycles of 1L platinum-doublet (bevacizumab allowed with chemotherapy) with a response were randomized 4:1 to oral rucaparib 600 mg BID or placebo. Pts were stratified by HRD status (as determined by FoundationOne CDx), residual disease status after chemotherapy, and timing of surgery. The primary endpoint of investigator-assessed PFS per RECIST was assessed in a step-down procedure first in the HRD population (BRCA mutant or BRCA wild-type/loss of heterozygosity [LOH] high carcinoma) and then in the intent-to-treat (ITT) population. Blinded independent central review (BICR)–assessed PFS was a stand-alone, secondary endpoint. PFS in BRCA mutant and HRD-negative pts (BRCA wild-type/LOH low) were exploratory endpoints. Results: As of Mar 23, 2022 (visit cutoff), 427 and 111 pts were randomized to rucaparib monotherapy or placebo (median time on treatment, 14.7 and 9.9 mo). PFS data are shown in the Table. Most common grade ≥3 TEAEs were anemia (rucaparib, 28.7% vs placebo, 0%), neutropenia (14.6% vs 0.9%), and ALT/AST increased (10.6% vs 0.9%). Rucaparib dose reduction, interruption, and discontinuation due to TEAEs occurred in 49.4%, 60.7%, and 11.8% of pts. Conclusions: Rucaparib monotherapy is effective as 1L maintenance with significant benefit vs placebo observed in the ITT and HRD populations, as well as the non-nested subgroup of pts without known HRD. Clinical trial information: NCT03522246. [Table: see text]

Proximal operator for the sorted [formula omitted] norm: Application to testing procedures based on SLOPE

A decade ago OSCAR was introduced as a penalized estimator where the penalty term, the sorted ℓ1 norm, allows to perform clustering selection. More recently, SLOPE was introduced as a penalized estimator controlling the False Discovery Rate (FDR) as soon as the hyper-parameter of the sorted ℓ1 norm is properly selected. For both, OSCAR and SLOPE, numerical schemes to compute these estimators are based on the proximal operator of the sorted ℓ1 norm. The main goal of this note is to provide a short and simple formula for this operator. Based on this formula one may observe that the output of the proximal operator has some components equal and thus this formula corroborates that SLOPE, as well as OSCAR, perform clustering selection. Moreover, our geometric approach to prove the formula for the proximal operator provides insights to show that testing procedures based on SLOPE are more powerful than step-down testing procedures but less powerful than step-up testing procedures.

A Definitive Prognostication System for Patients With Thoracic Malignancies Diagnosed With Coronavirus Disease 2019: An Update From the TERAVOLT Registry

IntroductionPatients with thoracic malignancies are at increased risk for mortality from coronavirus disease 2019 (COVID-19), and a large number of intertwined prognostic variables have been identified so far. MethodsCapitalizing data from the Thoracic Cancers International COVID-19 Collaboration (TERAVOLT) registry, a global study created with the aim of describing the impact of COVID-19 in patients with thoracic malignancies, we used a clustering approach, a fast-backward step-down selection procedure, and a tree-based model to screen and optimize a broad panel of demographics and clinical COVID-19 and cancer characteristics. ResultsAs of April 15, 2021, a total of 1491 consecutive eligible patients from 18 countries were included in the analysis. With a mean observation period of 42 days, 361 events were reported with an all-cause case fatality rate of 24.2%. The clustering procedure screened 73 covariates in 13 clusters. A further multivariable logistic regression for the association between clusters and death was performed, resulting in five clusters significantly associated with the outcome. The fast-backward step-down selection procedure then identified the following seven major determinants of death: Eastern Cooperative Oncology Group—performance status (ECOG-PS) (OR = 2.47, 1.87–3.26), neutrophil count (OR = 2.46, 1.76–3.44), serum procalcitonin (OR = 2.37, 1.64–3.43), development of pneumonia (OR = 1.95, 1.48–2.58), C-reactive protein (OR = 1.90, 1.43–2.51), tumor stage at COVID-19 diagnosis (OR = 1.97, 1.46–2.66), and age (OR = 1.71, 1.29–2.26). The receiver operating characteristic analysis for death of the selected model confirmed its diagnostic ability (area under the receiver operating curve = 0.78, 95% confidence interval: 0.75–0.81). The nomogram was able to classify the COVID-19 mortality in an interval ranging from 8% to 90%, and the tree-based model recognized ECOG-PS, neutrophil count, and c-reactive protein as the major determinants of prognosis. ConclusionsFrom 73 variables analyzed, seven major determinants of death have been identified. Poor ECOG-PS was found to have the strongest association with poor outcome from COVID-19. With our analysis, we provide clinicians with a definitive prognostication system to help determine the risk of mortality for patients with thoracic malignancies and COVID-19.

Feasibility of a Clinical-Radiomics Model to Predict the Outcomes of Acute Ischemic Stroke.

ObjectiveTo develop a model incorporating radiomic features and clinical factors to accurately predict acute ischemic stroke (AIS) outcomes.Materials and MethodsData from 522 AIS patients (382 male [73.2%]; mean age ± standard deviation, 58.9 ± 11.5 years) were randomly divided into the training (n = 311) and validation cohorts (n = 211). According to the modified Rankin Scale (mRS) at 6 months after hospital discharge, prognosis was dichotomized into good (mRS ≤ 2) and poor (mRS > 2); 1310 radiomics features were extracted from diffusion-weighted imaging and apparent diffusion coefficient maps. The minimum redundancy maximum relevance algorithm and the least absolute shrinkage and selection operator logistic regression method were implemented to select the features and establish a radiomics model. Univariable and multivariable logistic regression analyses were performed to identify the clinical factors and construct a clinical model. Ultimately, a multivariable logistic regression analysis incorporating independent clinical factors and radiomics score was implemented to establish the final combined prediction model using a backward step-down selection procedure, and a clinical-radiomics nomogram was developed. The models were evaluated using calibration, receiver operating characteristic (ROC), and decision curve analyses.ResultsAge, sex, stroke history, diabetes, baseline mRS, baseline National Institutes of Health Stroke Scale score, and radiomics score were independent predictors of AIS outcomes. The area under the ROC curve of the clinical-radiomics model was 0.868 (95% confidence interval, 0.825–0.910) in the training cohort and 0.890 (0.844–0.936) in the validation cohort, which was significantly larger than that of the clinical or radiomics models. The clinical radiomics nomogram was well calibrated (p > 0.05). The decision curve analysis indicated its clinical usefulness.ConclusionThe clinical-radiomics model outperformed individual clinical or radiomics models and achieved satisfactory performance in predicting AIS outcomes.

What's in Your Nomogram? Personalized Prognostication of Verbal Memory Decline after Temporal Lobe Resection in Adults With Epilepsy.

ObjectiveThis study aims to develop and externally validate models to predict the probability of postoperative verbal memory decline in adults following temporal lobe resection (TLR) for epilepsy using easily accessible preoperative clinical predictors.MethodsMultivariable models were developed to predict delayed verbal memory outcome on 3 commonly used measures: Rey Auditory Verbal Learning Test (RAVLT), and Logical Memory (LM), and Verbal Paired Associates (VPA) subtests from Wechsler Memory Scale-Third Edition. Using Harrell’s step-down procedure for variable selection, models were developed in 359 adults who underwent TLR at Cleveland Clinic and validated in 290 adults at 1 of 5 epilepsy surgery centers in the United States or Canada.ResultsTwenty-nine percent of the development cohort and 26% of the validation cohort demonstrated significant decline on at least 1 verbal memory measure. Initial models had good-to-excellent predictive accuracy (calibration (c) statistic range = .77–.80) in identifying patients with memory decline; however, models slightly underestimated decline in the validation cohort. Model coefficients were updated using data from both cohorts to improve stability. The model for RAVLT included surgery side, baseline memory score, and hippocampal resection. The models for LM and VPA included surgery side, baseline score, and education. Updated model performance was good to excellent (RAVLT c = .81, LM c = .76, VPA c = .78). Model calibration was very good, indicating no systematic over- or under-estimation of risk.ConclusionsNomograms are provided in 2 easy-to-use formats to assist clinicians in estimating the probability of verbal memory decline in adults considering TLR for treatment of epilepsy.

Step-Down Procedure for Comparison between Successive Exponential Populations

This article deals with the comparison between successive two-parameter exponential populations with respect to the location parameter when it is reasonable to assume that the populations are ordered in some natural way. It is well reported in the literature that if the interest of the experimenter is only in testing the significance of differences, the stepwise procedures are more powerful than the simultaneous confidence intervals procedures. This fact motivated us to extend the simultaneous confidence intervals procedure, for the differences between the location parameters of successive exponential populations, of Singh et al. (2006) to a stepwise procedure by proposing step-down tests for simultaneously testing the significance of the differences of the location parameters of successive two-parameter exponential populations. For a given type-I family-wise error rate (FWER), the critical constants are tabulated for the implementation of the proposed procedure by the practitioners. The advantage of the proposed procedure in comparison to the Singh et al. (2006) procedure is demonstrated by using a numerical example.

Differentially private false discovery rate control

Differential privacy provides a rigorous framework for privacy-preserving data analysis. This paper proposes the first differentially private procedure for controlling the false discovery rate (FDR) in multiple hypothesis testing. Inspired by the Benjamini-Hochberg procedure (BHq), our approach is to first repeatedly add noise to the logarithms of the p-values to ensure differential privacy and to select an approximately smallest p-value serving as a promising candidate at each iteration; the selected p-values are further supplied to the BHq and our private procedure releases only the rejected ones. Moreover, we develop a new technique that is based on a backward submartingale for proving FDR control of a broad class of multiple testing procedures, including our private procedure, and both the BHq step- up and step-down procedures. As a novel aspect, the proof works for arbitrary dependence between the true null and false null test statistics, while FDR control is maintained up to a small multiplicative factor.

Improving the prediction of epilepsy surgery outcomes using basic scalp EEG findings.

This study aims to evaluate the role of scalp electroencephalography (EEG; ictal and interictal patterns) in predicting resective epilepsy surgery outcomes. We use the data to further develop a nomogram to predict seizure freedom. We retrospectively reviewed the scalp EEG findings and clinical data of patients who underwent surgical resection at three epilepsy centers. Using both EEG and clinical variables categorized into 13 isolated candidate predictors and 6 interaction terms, we built a multivariable Cox proportional hazards model to predict seizure freedom 2years after surgery. Harrell's step-down procedure was used to sequentially eliminate the least-informative variables from the model until the change in the concordance index (c-index) with variable removal was less than 0.01. We created a separate model using only clinical variables. Discrimination of the two models was compared to evaluate the role of scalp EEG in seizure-freedom prediction. Four hundred seventy patient records were analyzed. Following internal validation, the full Clinical+EEG model achieved an optimism-corrected c-index of 0.65, whereas the c-index of the model without EEG data was 0.59. The presence of focal to bilateral tonic-clonic seizures (FBTCS), high preoperative seizure frequency, absence of hippocampal sclerosis, and presence of nonlocalizable seizures predicted worse outcome. The presence of FBTCS had the largest impact for predicting outcome. The analysis of the models' interactions showed that in patients with unilateral interictal epileptiform discharges (IEDs), temporal lobe surgery cases had a better outcome. In cases with bilateral IEDs, abnormal magnetic resonance imaging (MRI) predicted worse outcomes, and in cases without IEDs, patients with extratemporal epilepsy and abnormal MRI had better outcomes. This study highlights the value of scalp EEG, particularly the significance of IEDs, in predicting surgical outcome. The nomogram delivers an individualized prediction of postoperative outcome, and provides a unique assessment of the relationship between the outcome and preoperative findings.