Inverse Probability Of Censoring Weights Research Articles

Neurodevelopmental outcomes in children after prenatal marijuana exposure.

The effect of prenatal marijuana exposure (PME) on child neurodevelopment remains poorly understood. Prior studies have demonstrated inconsistent results. This study evaluated the association between PME and neuropsychological test scores in late childhood and early adulthood, accounting for a wide range of parental characteristics. This study evaluated participants from the Raine Study, a cohort of 2868 children born between 1989 and 1992. Children whose mothers provided information on marijuana use during pregnancy were included. The primary outcome was the Clinical Evaluation of Language Fundamentals (CELF) at age 10. Secondary outcomes included the Peabody Picture Vocabulary Test (PPVT), Child Behaviour Checklist (CBCL), McCarron Assessment of Neuromuscular Development (MAND), Coloured Progressive Matrices (CPM), Symbol Digit Modality Test (SDMT) and Autism Spectrum Quotient (AQ) scores. Exposed and unexposed children were matched by propensity score using optimal full matching. Missing covariate data were imputed using multiple imputation. Inverse probability of censoring weighting (IPCW) was used to adjust for missing outcome data. Linear regression within matched sets, adjusted by IPCW, evaluated score differences between exposed and unexposed children. As a secondary analysis, modified Poisson regression, adjusted by match weights and IPCW, evaluated the risk of clinical deficit in each outcome following PME. Of the 2804 children in this cohort, 285 (10.2%) had PME. After optimal full matching and IPCW, exposed children scored similarly on CELF Total (-0.33 points, 95% confidence interval [CI] -4.71, 4.05), Receptive (+0.65 points, 95% CI -4.08, 5.38) or Expressive (-0.53 points, 95% CI -5.07, 4.02). PME was not associated with secondary outcomes or risks of clinical deficit in any neuropsychological assessments. After adjusting for sociodemographic and clinical covariates, PME was not associated with worse neuropsychological test scores at age 10 or autistic traits at 19-20.

Real-world overall survival in second-line maintenance niraparib monotherapy vs active surveillance in BRCA wild-type patients with recurrent ovarian cancer.

5592 Background: The Phase 3NOVA trial assessed niraparib versus placebo as a maintenance treatment (tx) for patients (pts) with recurrent ovarian cancer (OC) who were platinum-sensitive after ≥2 regimens; progression-free survival was the primary endpoint and overall survival (OS) was a secondary endpoint (NCT01847274). The aim of this real-world study was to compare OS in a breast cancer gene wild-type ( BRCAwt) population of pts with recurrent OC and who received second-line maintenance (2Lm) niraparib monotherapy or were under active surveillance (AS). Methods: This study used the US nationwide Flatiron Health de-identified electronic health record (EHR)-derived database. Pts diagnosed with epithelial OC between 1Jan2011–31May2021 who completed 2L therapy between 1Jan2017–2Mar2022 were eligible for inclusion. Pts were BRCAwt, had an ECOG PS score 0–1, epithelial histology, and platinum-sensitive disease with ≥6 months between end of first-line tx and start of 2L tx. Pts were assigned to 2Lm niraparib or AS cohorts based on their management following 2L non-maintenance therapy (≤120 days). Follow-up was measured from the index date (end of 2L non-maintenance therapy) until end of study (31May2022), last activity, or death, whichever came first. A target trial emulation cloned inverse probability of censoring weighting (IPCW) methodology was selected a priori to minimize bias. IPCW median OS and hazard ratios (HR) for 2Lm versus AS were estimated with Kaplan-Meier curves and Cox regression models. Results: Overall, 266 BRCAwt pts were included and received niraparib 2Lm (N=123) or were under AS (N=143). Across 2Lm and AS cohorts, prior to cloning and IPCW, 23.6% and 34.3% of pts were aged ≥75 years, and 26.0% and 16.8% had an EHR value for race as other than White, respectively. The majority of pts were from community practice (80.5% and 90.2%, respectively), had serous OC (79.7% and 81.1%) and had Stage III disease (53.7% and 60.8%). Homologous recombination deficiency status was unknown for 92.7% and 86.0% of pts in 2Lm and AS cohorts, respectively. Median follow-up for 2Lm and AS cohorts was 16.8 (Quartile [Q] 1, Q3: 10.4, 28.7) and 10.2 (Q1, Q3: 4.1, 23.7) months, respectively. IPCW median OS was 28.1 (95% confidence interval [CI]: 22.5, 43.2) and 21.5 (95% CI: 14.7, 27.0) months (HR: 0.63 [95% CI: 0.45, 0.88]) for niraparib 2Lm and AS cohorts, respectively and survival rates at 24 months were 58.2% (95% CI: 47.5, 67.6) and 46.1% (95% CI: 33.6, 57.7). Conclusions: This analysis provides real-world data on OS among pts with OC receiving niraparib 2Lm or under AS in the 2L setting following chemotherapy. Funding: GSK study (219306). Editorial support was provided by Fishawack Health, funded by GSK.

Effects of [177Lu] Lu-PSMA-617 on overall survival in TheraP versus VISION randomized trials: An exploratory analysis.

5045 Background: Observed effects on overall survival (OS) differed in the first two randomized trials (RCTs) of [177Lu] Lu-PSMA-617 (LuPSMA) in metastatic, castration-resistant prostate cancer. TheraP randomized participants to LuPSMA or Cabazitaxel (CBZ). In TheraP, the CBZ group (n = 101) included 20 participants who crossed over to LuPSMA, and the LuPSMA group (n = 99) included 32 who crossed over to CBZ. VISION randomized participants to treatment with or without LuPSMA, plus physicians’ choice among protocol-defined treatments, excluding CBZ. Thirty-eight percent of control group participants in VISION had received CBZ prior to randomization. This study investigated explanations for differences in these RCTs in the observed effects of LuPSMA on OS. Methods: Individual participant data (IPD) on survival time from VISION was extracted from published survival curves for comparison with IPD from TheraP. Rank-preserving structural failure time (RPSFT) and inverse probability of censoring weighting (IPCW) were used in TheraP to estimate treatment effects on OS in counterfactual analyses assuming crossover had not occurred in either group, or in both groups. Results: OS was longer in the comparator group of TheraP than the comparator group of VISION (HR: 0.57, 95% CI 0.43 - 0.75) while OS was similar in the LuPSMA groups of VISION and TheraP (HR: 0.92, 95% CI 0.70 - 1.19). No significant differences in OS between the LuPSMA and CBZ groups of TheraP were apparent using counterfactual analyses assuming crossovers had not occurred (No crossover from CBZ: RPSFT HR 0.97, 95% CI 0.62 –1.52, IPCW HR 0.92, 95% CI 0.65-1.32; No crossover from LuPSMA: RPSFT HR 0.97, 95% CI 0.60–1.58, IPCW HR 0.82, 95% CI 0.55–1.24; No crossover from LuPSMA or CBZ: RPSFT HR 0.97, 95% CI 0.53-1.74, IPCW HR 0.82, 95% CI 0.53–1.27). Conclusions: The difference in observed effects of LuPSMA on overall survival in TheraP and VISION are better explained by the use of different comparator treatments in their control groups than by crossovers in TheraP from Cabazitaxel to LuPSMA, or LuPMSA to Cabazitaxel. Clinical trial information: NCT03392428 , NCT03511664 .

Impact of the treatment crossover design on comparative efficacy in EMPOWER-Lung 1: Cemiplimab monotherapy as first-line treatment of advanced non-small cell lung cancer.

In randomized-controlled crossover design trials, overall survival (OS) treatment effect estimates are often confounded by the control group benefiting from treatment received post-progression. We estimated the adjusted OS treatment effect in EMPOWER-Lung 1 (NCT03088540) by accounting for the potential impact of crossover to cemiplimab among controls and continued cemiplimab treatment post-progression. Patients were randomly assigned 1:1 to cemiplimab 350 mg every 3 weeks (Q3W) or platinum-doublet chemotherapy. Patients with disease progression while on or after chemotherapy could receive cemiplimab 350 mg Q3W for ≤108 weeks. Those who experienced progression on cemiplimab could continue cemiplimab at 350 mg Q3W for ≤108 additional weeks with four chemotherapy cycles added. Three adjustment methods accounted for crossover and/or continued treatment: simplified two-stage correction (with or without recensoring), inverse probability of censoring weighting (IPCW), and rank-preserving structural failure time model (RPSFT; with or without recensoring). In the programmed cell death-ligand 1 ≥50% population (N=563; median 10.8-month follow-up), 38.2% (n=107/280) crossed over from chemotherapy to cemiplimab (71.3%, n=107/150, among those with confirmed progression) and 16.3% (n=46/283) received cemiplimab treatment after progression with the addition of histology-specific chemotherapy (38.7%, n=46/119, among those with confirmed progression). The unadjusted OS hazard ratio (HR) with cemiplimab versus chemotherapy was 0.566 (95% confidence interval [CI]: 0.418, 0.767). Simplified two-stage correction-the most suitable method based on published guidelines and trial characteristics-produced an OS HR of 0.490 (95% CI: 0.365, 0.654) without recensoring and 0.493 (95% CI: 0.361, 0.674) with recensoring. The IPCW and RPSFT methods produced estimates generally consistent with simplified two-stage correction. After adjusting for treatment crossover and continued cemiplimab treatment after progression with the addition of histology-specific chemotherapy observed in EMPOWER-Lung 1, cemiplimab continued to demonstrate a clinically important and statistically significant OS benefit versus chemotherapy, consistent with the primary analysis.

IPCW approach for testing independence

Here we present a novel inverse probability of censoring weighted (IPCW) adaptation of the Kochar–Gupta (KG) test of independence, in the case of bivariate randomly censored data. Three different censoring schemes are considered: one of the target variables is censored, both targeted variables are censored with the same censoring variable, and both target variables are censored with different censoring variables. The limiting properties of test statistics are explored. In order to compare the tests with a few well-known competitors in terms of powers, several resampling procedures have been utilised to approximate the null distribution. Special attention is dedicated to the comparison with a classical adaptation of the KG test related to the IPCW adaptation of U-statistics.

Age-Dependent Hemorrhage Risk and Obliteration Benefit After Radiosurgery for Brain Arteriovenous Malformation

Because patients with untreated brain arteriovenous malformations (BAVMs) are at variable risks of cerebral hemorrhage and associated mortality and morbidity, it is essential to identify patient populations who benefit most from prophylactic interventions. This study aimed to examine age-dependent differences in the therapeutic effect of stereotactic radiosurgery (SRS) on BAVMs. This retrospective observational study enrolled patients with BAVMs who underwent SRS at our institution between 1990 and 2017. The primary outcome was post-SRS hemorrhage, and the secondary outcomes included nidus obliteration, post-SRS early signal changes, and mortality. To investigate age-related differences in outcomes after SRS, we performed age-stratified analyses using the Kaplan-Meier analysis and weighted logistic regression with the inverse probability of censoring weighting (IPCW). To address significant differences in patient baseline characteristics, we also performed inverse probability of treatment weighting (IPTW) adjusted for possible confounders to investigate age-related differences in outcomes after SRS. A total of 735 patients with 738 BAVMs were stratified by age. Age-stratified analysis using a weighted logistic regression model with IPCW showed a direct correlation between patient age and post-SRS hemorrhage (odds ratio [OR], 95% confidence interval [CI], and P value: 2.20, 1.34-3.63, and .002 at 18 months; 1.86, 1.17-2.93, and .008 at 36 months; and 1.61, 1.05-2.48, and .030 at 54 months, respectively). The age-stratified analysis also showed an inverse relationship between age and obliteration over the first 42 months after SRS (OR, 95% CI, and P value: 0.05, 0.02-0.12, and <.001 at 6 months; 0.55, 0.44-0.70, and <.001 at 24 months; and 0.76, 0.63-0.91, and .002 at 42 months, respectively). These results were also confirmed with the IPTW analyses. Our analysis demonstrated that patients' age at SRS is significantly associated with hemorrhage and the nidus obliteration rate after treatment. In particular, younger patients are more likely to exhibit reduced cerebral hemorrhages and achieve earlier nidus obliteration compared with older patients.

Informative Censoring-A Cause of Bias in Estimating COVID-19 Mortality Using Hospital Data.

(1) Background: Several retrospective observational analyzed treatment outcomes for COVID-19; (2) Methods: Inverse probability of censoring weighting (IPCW) was applied to correct for bias due to informative censoring in database of hospitalized patients who did and did not receive convalescent plasma; (3) Results: When compared with an IPCW analysis, overall mortality was overestimated using an unadjusted Kaplan-Meier curve, and hazard ratios for the older age group compared to the youngest were underestimated using the Cox proportional hazard models and 30-day mortality; (4) Conclusions: An IPCW analysis provided stabilizing weights by hospital admission.

Uniform Consistency for Functional Conditional U-Statistics Using Delta-Sequences

U-statistics are a fundamental class of statistics derived from modeling quantities of interest characterized by responses from multiple subjects. U-statistics make generalizations the empirical mean of a random variable X to the sum of all k-tuples of X observations. This paper examines a setting for nonparametric statistical curve estimation based on an infinite-dimensional covariate, including Stute’s estimator as a special case. In this functional context, the class of “delta sequence estimators” is defined and discussed. The orthogonal series method and the histogram method are both included in this class. We achieve almost complete uniform convergence with the rates of these estimators under certain broad conditions. Moreover, in the same context, we show the uniform almost-complete convergence for the nonparametric inverse probability of censoring weighted (I.P.C.W.) estimators of the regression function under random censorship, which is of its own interest. Among the potential applications are discrimination problems, metric learning and the time series prediction from the continuous set of past values.

Model selection for survival individualized treatment rules using the jackknife estimator

BackgroundPrecision medicine is an emerging field that involves the selection of treatments based on patients’ individual prognostic data. It is formalized through the identification of individualized treatment rules (ITRs) that maximize a clinical outcome. When the type of outcome is time-to-event, the correct handling of censoring is crucial for estimating reliable optimal ITRs.MethodsWe propose a jackknife estimator of the value function to allow for right-censored data for a binary treatment. The jackknife estimator or leave-one-out-cross-validation approach can be used to estimate the value function and select optimal ITRs using existing machine learning methods. We address the issue of censoring in survival data by introducing an inverse probability of censoring weighted (IPCW) adjustment in the expression of the jackknife estimator of the value function. In this paper, we estimate the optimal ITR by using random survival forest (RSF) and Cox proportional hazards model (COX). We use a Z-test to compare the optimal ITRs learned by RSF and COX with the zero-order model (or one-size-fits-all). Through simulation studies, we investigate the asymptotic properties and the performance of our proposed estimator under different censoring rates. We illustrate our proposed method on a phase III clinical trial of non-small cell lung cancer data.ResultsOur simulations show that COX outperforms RSF for small sample sizes. As sample sizes increase, the performance of RSF improves, in particular when the expected log failure time is not linear in the covariates. The estimator is fairly normally distributed across different combinations of simulation scenarios and censoring rates. When applied to a non-small-cell lung cancer data set, our method determines the zero-order model (ZOM) as the best performing model. This finding highlights the possibility that tailoring may not be needed for this cancer data set.ConclusionThe jackknife approach for estimating the value function in the presence of right-censored data shows satisfactory performance when there is small to moderate censoring. Winsorizing the upper and lower percentiles of the estimated survival weights for computing the IPCWs stabilizes the estimator.

Brentuximab Vedotin Plus CHP in Frontline sALCL: Adjusted Estimates of Efficacy and Cost-Effectiveness Removing the Effects of Re-Treatment with Brentuximab Vedotin.

BackgroundIn the randomised controlled trial ECHELON-2 (NCT01777152; January 2013), brentuximab vedotin (BV) plus cyclophosphamide, doxorubicin and prednisone (CHP) demonstrated improved efficacy compared with CHOP (CHP and vincristine) in frontline CD30+ peripheral T-cell lymphoma (PTCL), an aggressive cancer with poor survival. In ECHELON-2, 70% of patients had systemic anaplastic large cell lymphoma (sALCL), a subtype of PTCL. Of sALCL patients who progressed from BV+CHP and CHOP, 36% (n = 17) and 56% (n = 36) received subsequent BV-containing therapy, respectively. As BV re-treatment was not funded in England at the time, our objective was to estimate adjusted efficacy and cost-effectiveness by excluding BV re-treatment from BV+CHP.MethodsTo remove the effects of BV re-treatment, the inverse probability of censoring weights (IPCW) and two-stage estimator (TSE) approaches, with and without re-censoring, were applied to overall survival (OS) in the BV+CHP arm of the ECHELON-2 sALCL population. Cost-effectiveness was determined in a three-state partitioned survival (PartSA) model from the perspective of the National Health Service (NHS) in England.ResultsThe unadjusted hazard ratio (HR) for death in patients with sALCL with BV+CHP versus CHOP was 0.54 (95% CI 0.34, 0.87; p = 0.011). The model base case used TSE analysis without re-censoring, which provided an adjusted HR for death of 0.55 (95% CI 0.33, 0.86; p = 0.014). Incremental cost-effectiveness ratios (ICERs) including and excluding re-treatment with BV were £29,760/QALY and £27,761/QALY, respectively.ConclusionTSE without re-censoring provided the most clinically plausible estimate of survival whilst retaining sufficient information for OS extrapolation. After adjustment for BV re-treatment, BV+CHP remains an efficacious and cost-effective treatment in frontline sALCL compared with CHOP.Supplementary InformationThe online version contains supplementary material available at 10.1007/s41669-022-00349-z.

1026P Treatment switching adjustment of overall survival in the CheckMate 227 clinical trial of nivolumab plus ipilimumab versus chemotherapy in first-line treatment of patients with advanced non-small cell lung cancer

CheckMate 227 is a phase III clinical trial designed to evaluate the efficacy of nivolumab plus ipilimumab vs platinum doublet chemotherapy in patients with stage IV or recurrent non-small cell lung cancer (NSCLC) previously untreated for advanced disease. Some patients in the chemotherapy arm subsequently switched to an immunotherapy due to toxicity or progressive disease. We update overall survival estimates by adjusting for potential biases introduced by treatment switching from chemotherapy to subsequent immunotherapy. The Inverse Probability of Censoring Weights (IPCW) method that adjusts the treatment effect estimates in the presence of informative censoring was applied to adjust OS for treatment switching in the ITT population as a post hoc analysis based on 3-year follow up data. The IPCW method, being well recognized by health authorities, was considered as the base case. Other treatment switching adjustment methods were explored as sensitivity analysis. Exploratory analyses were also conducted in select subgroups. In part 1 of the trial, a total of 1166 patients (583 each arm) were included in the analysis. 235 (40.3%) patients in the chemotherapy arm switched to an immunotherapy. Median OS was 17.1 months (95% CI: 15.5 to 20.0) for the nivolumab plus ipilimumab arm and 14.0 months (95% CI: 12.6-15.4 months) for the chemotherapy arm. After adjustment for treatment switching, the median OS for the chemotherapy arm was 10.8 months (95% CI: 9.2 -12.8). The hazard ratio (HR) of nivolumab plus ipilimumab versus chemotherapy was 0.74 (95% CI: 0.65 to 0.85), which after adjustment reduced to 0.62 (95% CI: 0.53 to 0.73). Based on 3-year follow up data from CheckMate 227, adjusting for treatment switching biases in the chemotherapy arm resulted in a numerically greater estimated relative OS benefit with nivolumab plus ipilimumab over chemotherapy in previously untreated patients, with advanced NSCLC without EGFR/ALK aberrations than in the unadjusted analysis.

On logistic regression with right censored data, with or without competing risks, and its use for estimating treatment effects.

Simple logistic regression can be adapted to deal with right-censoring by inverse probability of censoring weighting (IPCW). We here compare two such IPCW approaches, one based on weighting the outcome, the other based on weighting the estimating equations. We study the large sample properties of the two approaches and show that which of the two weighting methods is the most efficient depends on the censoring distribution. We show by theoretical computations that the methods can be surprisingly different in realistic settings. We further show how to use the two weighting approaches for logistic regression to estimate causal treatment effects, for both observational studies and randomized clinical trials (RCT). Several estimators for observational studies are compared and we present an application to registry data. We also revisit interesting robustness properties of logistic regression in the context of RCTs, with a particular focus on the IPCW weighting. We find that these robustness properties still hold when the censoring weights are correctly specified, but not necessarily otherwise.

Restricted mean survival time regression model with time-dependent covariates.

In clinical or epidemiological follow‐up studies, methods based on time scale indicators such as the restricted mean survival time (RMST) have been developed to some extent. Compared with traditional hazard rate indicator system methods, the RMST is easier to interpret and does not require the proportional hazard assumption. To date, regression models based on the RMST are indirect or direct models of the RMST and baseline covariates. However, time‐dependent covariates are becoming increasingly common in follow‐up studies. Based on the inverse probability of censoring weighting (IPCW) method, we developed a regression model of the RMST and time‐dependent covariates. Through Monte Carlo simulation, we verified the estimation performance of the regression parameters of the proposed model. Compared with the time‐dependent Cox model and the fixed (baseline) covariate RMST model, the time‐dependent RMST model has a better prediction ability. Finally, an example of heart transplantation was used to verify the above conclusions.

Bias and Type I error Control in Correcting Treatment Effect for Treatment Switching Using Marginal Structural Models in Phase III Oncology Trials

ABSTRACT This research focuses on the bias and type I error control issues when the marginal structural models (MSMs) are applied to evaluate the causal survival benefits of active intervention versus control in randomized clinical trials (RCTs) with treatment switching after disease progression. When MSMs are applied in the RCT setting, the question of interest, model specifications, strategies for type I error control, bias reduction, etc. differ somewhat from those for observational studies. This manuscript discusses the approaches used to accommodate these differences. Through Monte Carlo simulations and a case study, our research demonstrates that, with sufficient attention paid to issues applicable to RCTs in particular, MSMs may perform better than the inverse probability of censoring weighting (IPCW) method in analyzing the survival endpoint in RCTs with treatment switching because more information is used by the MSM.

Using observational study data as an external control group for a clinical trial: an empirical comparison of methods to account for longitudinal missing data

BackgroundObservational data are increasingly being used to conduct external comparisons to clinical trials. In this study, we empirically examined whether different methodological approaches to longitudinal missing data affected study conclusions in this setting.MethodsWe used data from one clinical trial and one prospective observational study, both Norwegian multicenter studies including patients with recently diagnosed rheumatoid arthritis and implementing similar treatment strategies, but with different stringency. A binary disease remission status was defined at 6, 12, and 24 months in both studies. After identifying patterns of longitudinal missing outcome data, we evaluated the following five approaches to handle missingness: analyses of patients with complete follow-up data, multiple imputation (MI), inverse probability of censoring weighting (IPCW), and two combinations of MI and IPCW.ResultsWe found a complex non-monotone missing data pattern in the observational study (N = 328), while missing data in the trial (N = 188) was monotone due to drop-out. In the observational study, only 39.0% of patients had complete outcome data, compared to 89.9% in the trial. All approaches to missing data indicated favorable outcomes of the treatment strategy in the trial and resulted in similar study conclusions. Variations in results across approaches were mainly due to variations in estimated outcomes for the observational data.ConclusionsFive different approaches to handle longitudinal missing data resulted in similar conclusions in our example. However, the extent and complexity of missing observational data affected estimated comparative outcomes across approaches, highlighting the need for careful consideration of methods to account for missingness in this setting. Based on this empirical examination, we recommend using a prespecified advanced missing data approach to account for longitudinal missing data, and to conduct alternative approaches in sensitivity analyses.

Survival stacking with multiple data types using pseudo-observation-based-AUC loss

ABSTRACT There have been many strategies to adapt machine learning algorithms to account for right censored observations in survival data in order to build more accurate risk prediction models. These adaptions have included pre-processing steps such as pseudo-observation transformation of the survival outcome or inverse probability of censoring weighted (IPCW) bootstrapping of the observed binary indicator of an event prior to a time point of interest. These pre-processing steps allow existing or newly developed machine learning methods, which were not specifically developed with time-to-event data in mind, to be applied to right censored survival data for predicting the risk of experiencing an event. Stacking or ensemble methods can improve on risk predictions, but in general, the combination of pseudo-observation-based algorithms, IPCW bootstrapping, IPC weighting of the methods directly, and methods developed specifically for survival has not been considered in the same ensemble. In this paper, we propose an ensemble procedure based on the area under the pseudo-observation-based-time-dependent ROC curve to optimally stack predictions from any survival or survival adapted algorithm. The real application results show that our proposed method can improve on single survival based methods such as survival random forest or on other strategies that use a pre-processing step such as inverse probability of censoring weighted bagging or pseudo-observations alone.

Comparative safety and effectiveness of ticagrelor versus clopidogrel in patients with acute coronary syndrome. An on treatment analysis from the multicenter CREA ARIAM Andalucia registry

Abstract Funding Acknowledgements Type of funding sources: Private company. Main funding source(s): This study was financially by a non conditional grant by Astrazeneca. Background/Introduction The favorable net clinical benefit of ticagrelor over clopidogrel in the PLATO trial has been recently questioned in observational studies. However, these mostly retrospective analyses based on the intention-to-treat (ITT) principle, systematically failed to account for switching between P2Y12 inhibitors, which may indeed have led to biased estimates due to exposure misclassification (incorrect classification of subjects with respect to exposure) Purpose To assess the on-treatment safety and effectiveness of ticagrelor vs clopidogrel among patients with acute coronary syndrome (ACS) in a non-trial scenario. Methods Prospective multicentre cohort study of 2070 ACS patients discharged on ticagrelor or clopidogrel between 2015 and 2019. Major exclusions were previous intracranial bleeding, use of prasugrel or oral anticoagulation. Primary study hypothesis stated safety profile of ticagrelor is not unacceptably worse (not inferior) than that of clopidogrel. Primary outcome was risk of major bleeding (BARC types 3, 5) at 1 year. Secondary outcomes were major adverse cardiac and cerebrovascular events (MACCE), and net adverse clinical event (NACE) at 1-yer. Association of ticagrelor vs clopidogrel with outcomes was evaluated based on the on-treatment principle using fully-adjusted Cox regression models with double robust inverse probability of censoring weighted (IPCW) estimators. Sensitivity analyses included propensity score matching analysis and ITT simulation analysis. Results Among 2070 patients included (mean age 63 years, 73% men), 1035 were discharged on ticagrelor and clopidogrel in each treatment group, respectively. Ticagrelor-treated patients were younger, had less comorbidities and more often underwent percutaneous coronary intervention over coronary artery bypass grafting surgery, as the preferred revascularization choice. After adjust for medication adherence and switching, ticagrelor did not increase the risk for major bleeding compared with clopidogrel (adjusted hazard ratio 1.40, 95% CI 0.96 – 2.05; P = .075). Although there was signal for potential harm with ticagrelor among elderly people and patients with previous bleeding, exploratory analysis suggested consistent reductions in MACCE and NACE risks compared with clopidogrel. PSM analysis was congruent with primary analysis, whereas ITT analysis yielded results in the opposite direction. Conclusion(s) In this on-treatment analysis of an all-comers ACS population, ticagrelor did not pose an increased risk for major bleeding, while resulted in a net clinical benefit compared with clopidogrel. Further studies accounting for drug switching are warranted to confirm these findings in high-bleeding risk populations.

Machine Learning vs. Survival Analysis Models: a study on right censored heart failure data

Machine Learning Models are known to understand the intricacies of the data well, but native ML models cannot be used in time-to-event analysis due to censoring. In this paper, we explore the use of Machine Learning Models in the field of Survival Analysis using right censored Heart Failure Clinical Records Dataset. For this purpose, we first identify the top most important features responsible for death due to heart failure using Recursive Feature Elimination and then see how Machine Learning models can be adapted to improve the time-to-event analysis outcomes. To deal with this, Machine Learning Models are modified using the techniques Inverse Probability of Censoring Weighting (IPCW) and IPCW Bagging and are trained using the processed dataset alongside various survival analysis models. Area Under the time-dependent ROC (AUC) is used as a performance metric. The results reveal that the average AUC value for Survival Analysis Models is 0.51 while that of Machine Learning Models processed using IPCW increased to 0.80, and those processed using IPCW Bagging increased by 0.82. This reflects that Machine Learning models outperform Survival Analysis models in the case of time-to-event analysis of right censored dataset, and hence, are better indicators of risk of heart disease.

Estimation of Overall Survival with Subsequent Treatment Effect by Applying Inverse Probability of Censoring Weighting in the LATITUDE Study

BackgroundIn the LATITUDE study (ClinicalTrials.gov, NCT01715285), compared with placebos, abiraterone acetate plus prednisone (AAP) with androgen deprivation therapy (ADT) provided significant overall survival (OS) benefit in high-risk metastatic castration-sensitive prostate cancer (mCSPC) patients. It is controversial whether survival benefits would remain if all patients in the placebo group subsequently received life-extending therapies. ObjectiveTo estimate treatment effect in the case of all patients in the placebo group receiving life-extending subsequent therapies. Design, setting, and participantsA post hoc analysis of LATITUDE final-analysis data was carried out (setting and participants have been reported previously). InterventionAAP or placebos plus ADT. Outcome measurements and statistical analysisWe applied the inverse probability of censoring weighting (IPCW) method to represent the situation in which all patients in the placebo group would have received life-extending subsequent therapies. The OS hazard ratio (HR) of AAP versus placebos and associated 95% confidence interval (CI) were estimated using a Cox proportional hazards model. Results and limitationsOf the 581 eligible patients in the placebo group, 237 (40.8%) did not receive life-extending subsequent therapies. From the unadjusted intention-to-treat analysis, the HR for OS for AAP versus placebos was 0.661 (95% CI 0.564–0.775). Using IPCW to adjust for patients in the placebo group without life-extending subsequent therapies, the HR was 0.732 (95% CI 0.604–0.887). A limitation is a lack of proof that the Cox proportional hazards model for the absence of life-extending subsequent therapy is correctly specified for the IPCW method. ConclusionsTreatment with AAP exerts OS benefit over placebos in high-risk mCSPC patients, regardless of whether life-extending subsequent therapy is given. Patient summaryIn a previous study, high-risk metastatic castration-sensitive prostate cancer patients who received abiraterone acetate plus prednisone (AAP) with androgen deprivation therapy generally survived longer than those given placebos. The benefit of adding AAP continues regardless of whether life-extending subsequent therapy is given.

A Novel Algorithm for Analysis of Multiple Endpoints Using Risk–Benefit Profiles

Often it is necessary to evaluate effectiveness of an intervention on the basis of multiple event outcomes of variable benefit and harm, which may develop over time. An attractive approach is to order combinations of these events based on desirability of the overall outcome (e.g. from cure without any adverse events to death), and then determine whether the intervention shifts the distribution of these ordered outcomes towards more desirable (Evans, Follmann 2016). The win ratio introduced in Pocock et al 2012 was an earlier implementation of this approach. More recently Claggett et al 2015 proposed a more comprehensive method allowing nonparametric and regression-based inference in presence of competing risks. Key to the method is weighting observations by inverse probability of censoring (IPC) processes specific to participants and event types. The method has seemingly great practical utility, but computation of weights is a non-trivial challenge with real-life data when each event can have its own censoring time. We present a novel recursive algorithm solving this problem for an arbitrary number of events ordered by clinical importance or desirability. The algorithm can be implemented in SAS or R software, and computes IPC weights, as well as nonparametric or parametric estimates and resampling-based measures of uncertainty. We illustrate the approach using data from the SPRINT trial of antihypertensive intervention, comparing risk-benefit profiles for robust, pre-frail, and frail subpopulations, and in analysis of fall as a function of progressive risk factors. More general use of the software tools deploying the method is described.