General Information
Conflict of Interest
Request Clinical Trials
Associated Trial(s):- NCT02489318 - A Phase 3 Randomized, Placebo-controlled, Double-blind Study of Apalutamide Plus Androgen Deprivation Therapy (ADT) Versus ADT in Subjects With Metastatic Hormone-sensitive Prostate Cancer (mHSPC)
- NCT02257736 - A Phase 3 Randomized, Placebo-controlled Double-blind Study of JNJ-56021927 in Combination With Abiraterone Acetate and Prednisone Versus Abiraterone Acetate and Prednisone in Subjects With Chemotherapy-naive Metastatic Castration-resistant Prostate Cancer (mCRPC)
- NCT00887198 - A Phase 3, Randomized, Double-blind, Placebo-Controlled Study of Abiraterone Acetate (CB7630) Plus Prednisone in Asymptomatic or Mildly Symptomatic Patients With Metastatic Castration-Resistant Prostate Cancer
- NCT00638690 - A Phase 3, Randomized, Double-Blind, Placebo-Controlled Study of Abiraterone Acetate (CB7630) Plus Prednisone in Patients With Metastatic Castration-Resistant Prostate Cancer Who Have Failed Docetaxel-Based Chemotherapy
Request Clinical Trials
Data Request Status
Status: OngoingResearch Proposal
Project Title: Detection of Subgroup Treatment Effects with Missing Patient-Reported Outcomes Data
Scientific Abstract:
1.Background
Patient-reported outcome (PRO) measures are included in clinical trials to provide insights into patients' perspectives on treatment. The treatment effect can vary across subjects, which is defined as the individualized treatment effect (ITE), and are important in the decision making about clinical treatment. However, when it comes to the PRO data, it is challenging due to the skewness of responses and the presence of missing data. Current methods ignore these issues or only estimate average effects, not individualized benefits.
2.Objective
To propose a novel method to estimate the ITEs and detect patients with or without significant treatment effect by longitudinal PRO data with missingness.
3.Study Design
We have examined the efficiency of our proposed method in numerical experiments. We will apply the proposed method on the trial data to show the application of our method.
4.Participants
The study population in each trial.
5.Primary and Secondary Outcome Measure(s)
The Functional Assessment of Cancer Therapy -- Prostate (FACT-P) score.
6.Statistical Analysis
We will use the trial data to estimate the ITE of each patient and detect the subgroup of patients with or without significant treatment effect. First, we model the ITE as a non-linear covariate specific treatment effect (CSTE) curve in a semi-parametric quantile model. Then to address the missingness, we use an inverse probability (IPW) method and estimate the missing probability by a logistic model. The within-subject correlation is handled by a weighting method borrowing information from a mean model.
Brief Project Background and Statement of Project Significance:
Patient-Reported Outcomes (PROs) are increasingly assessed in clinical trials as primary, secondary,
or exploratory endpoints, providing the patients' perspective of treatment and the impact of medical conditions (Mercieca-Bebber et al., 2018). The value of PRO data has been highlighted by regulatory bodies, as an important factor when assessing the economic, social, and ethical implications of the approval and use of a treatment.
Traditional analyses mostly considered the average treatment effect and the subgroup analysis examines whether treatment effects vary across specific patient subgroups defined by characteristics such as age, gender, or disease severity. However, the division of patient subgroup may lose information from the data, for the artificial dichotomisation of continuous covariates. It could also reduce statistical power and an increased risk of false-positive findings due to multiple comparisons. Consequently, some individuals may not experience the anticipated benefits from a treatment that appears effective on average.
This limitation underscores the need for methodologies capable of assessing heterogeneity in treatment effects and tailor treatments more effectively to individual patients. Biomedical researches have paid much attention on the topic of "personalized medicine", which aims to recommend individualize treatment regime (ITR) according to patients' unique characteristics (Kent et al., 2018; Kosorok and Laber, 2019). A popular method to estimate the ITR is to apply non-parametric and semi-parametric models on the difference in average outcomes between two treatment groups conditionally on predictive characteristics (Ma and Zhou, 2014; Han et al., 2017; Guo et al. 2021). However, existing methods have some drawbacks in analysis on the PRO data. Usually, the range of PRO scores is restricted by the minimum and maximum scores of questions, and thus, in general, the PRO scores are not normal. As an alternative to the mean regression, quantile regression (Koenker, 2005) has been extended to the analysis of data when the distribution of the response is quite different from a normal distribution. Recently, Chatla and Bhattacharya (2025) extended the model proposed in Guo et al. (2021) to a quantile regression model for estimation and inference of individualized quantile treatment effect.
We aim to propose a novel semi-parametric quantile model to estimate the individualized treatment effect (ITE) using PRO data which are longitudinal with monotone missing responses. We adopted the IPW method to reduce the bias arising from missingness in responses and borrow information from a mean model to handle the within-subject correlation induced by the longitudinal structure. The proposed method can estimated the ITE on the PRO data by which the patients can be allocated to the optimal treatment group and get more benefits from different treatments.
Specific Aims of the Project:
Specific hypotheses to be evaluated:
Some characteristics are related to treatment effect, and some patients with some levels of characteristic will benefit from a new treatment while some will not.
The primary objective is to apply our proposed method to estimate the ITEs with longitudinal PRO data with missingness. We would like to emphasize that this analysis is intended to demonstrate the applicability of our method rather than to draw any medical conclusions from the results.
Study Design: Methodological research
What is the purpose of the analysis being proposed? Please select all that apply.: Develop or refine statistical methods Research on clinical trial methods Research on clinical prediction or risk prediction
Software Used: RStudio
Data Source and Inclusion/Exclusion Criteria to be used to define the patient sample for your study:
NCT02257736, NCT02489318, NCT00887198 and NCT00638690 are requested for this study. The IPD data from both trial will be used in its entirety.
There are no exclusion criteria defined.
Primary and Secondary Outcome Measure(s) and how they will be categorized/defined for your study: The primary outcome measure will be the Functional Assessment of Cancer Therapy -- Prostate (FACT-P), with the definition consistent with that used in the trials (NCT02257736, NCT02489318, NCT00887198 and NCT00638690).
Main Predictor/Independent Variable and how it will be categorized/defined for your study: The main predictor is the treatment arm that patients allocated and those may be related to the severity of diseases, or to the prognosis, which are more likely to be related to the treatment effect (e.g. the ECOG score, the tumor stage and the PSA value). All of these variables will be defined as the original trials.
Other Variables of Interest that will be used in your analysis and how they will be categorized/defined for your study: Baseline demographical characteristics like age, gender, weight and BMI. All of these variables will be defined as the original trials.
Statistical Analysis Plan:
Step 1: First, we will do a descriptive analysis of the trial data, giving a general picture of the missing patterns in the longitudinal observations. Specifically, we will summarize the dropout rate in each group by visits, and the percentage of subjects who have different numbers of missing items at each visit. We will also examine missing data in baseline covariates, since missing covariates is not the focus of this study, we will only include those covariates with very few missing data for later analysis.
Step 2: Conduct exploratory analysis to find out the potential variables most related to the interested outcome.
Step 3: Fit a logistic regression model to estimate a weight related to missingness per subject per visit.
Step 4: Estimate the empirical likelihood to handle the within-subject correlation.
Step 5, Fit a weighted quantile semi-parametric model to estimate the ITE at different levels of quantile.
Step 6: Identify subgroups of patients who may benefit from the treatment.
Narrative Summary: Patient-reported outcomes (PROs) track treatment impact but are hard to analyze due to skewed data, missing responses, and repeated measurements over time. Current methods ignore these PRO specific issues or only estimate average effects instead of individualized benefits, which is the core of precision medicine. We develop a new statistical method to estimate personalized treatment effects (ITEs) from longitudinal PROs with missing data, focusing on quantiles (e.g., median) rather than averages. This method will make better identification of patient subgroups that benefit (or don't) from treatments, improving decision-making for precision medicine.
Project Timeline: Our research has already begun and we have completed the design of model and numerical experiments. Manuscript writing aside from the real-data component will be completed by May 2025, and analysis on the trial data as well as the finalization of the manuscript is estimated to be completed by June 2025. the manuscript will be submitted for publication by August 2025.
Dissemination Plan: The manuscript will be submitted to journals in the field of statistical methods, such us Statistics in Medicine.
Bibliography:
Chatla, S. B. and Bhattacharya, I. (2025). Estimation and inference of quantile treatment effects in high-dimensional single-index model. Econometrics and Statistics.
Guo, W., Zhou, X.-H., and Ma, S. (2021). Estimation of optimal individualized treatment rules using a covariate-specific treatment effect curve with high-dimensional covariates. Journal of the American Statistical Association, 116(533):309--321.
Han, K., Zhou, X., and Liu, B. (2017). Cste curve for selection the optimal treatment when outcome is binary. Sci China Math, 47:497--514.
Ma, Y. and Zhou, X.-H. (2014). Treatment selection in a randomized clinical trial via covariate-specifictreatment effect curves. Statistical methods in medical research, 26(1):124--141
Mercieca-Bebber, R., King, M. T., Calvert, M. J., Stockler, M. R., and Friedlander, M. (2018). The importance of patient-reported outcomes in clinical trials and strategies for future optimization. Patient related outcome measures, pages 353--367.
Kent, D. M., Steyerberg, E., and Van Klaveren, D. (2018). Personalized evidence based medicine: predictive approaches to heterogeneous treatment effects. Bmj, 363.
Koenker, R. (2005). Quantile regression, volume 38. Cambridge university press.