array(41) {
  ["project_status"]=>
  string(7) "ongoing"
  ["project_assoc_trials"]=>
  array(6) {
    [0]=>
    object(WP_Post)#5543 (24) {
      ["ID"]=>
      int(1633)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2017-10-13 14:33:00"
      ["post_date_gmt"]=>
      string(19) "2017-10-13 14:33:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(322) "A group comparative, placebo-controlled, double-blind trial of the efficacy and safety of galantamine hydrobromide, 7.5 mg (6 mg galantamine base) TID, 10 mg (8 mg galantamine base) TID and 15 mg (12 mg galantamine base) TID taken orally for 12 weeks in patients with a diagnosis of senile dementia of the Alzheimer's type"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(190) "a-group-comparative-placebo-controlled-double-blind-trial-of-the-efficacy-and-safety-of-galantamine-hydrobromide-7-5-mg-6-mg-galantamine-base-tid-10-mg-8-mg-galantamine-base-tid-and-15-mg-12"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2026-01-07 10:38:10"
      ["post_modified_gmt"]=>
      string(19) "2026-01-07 15:38:10"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(239) "https://dev-yoda.pantheonsite.io/clinical-trial/a-group-comparative-placebo-controlled-double-blind-trial-of-the-efficacy-and-safety-of-galantamine-hydrobromide-7-5-mg-6-mg-galantamine-base-tid-10-mg-8-mg-galantamine-base-tid-and-15-mg-12/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
    [1]=>
    object(WP_Post)#5544 (24) {
      ["ID"]=>
      int(1552)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2016-08-03 14:04:00"
      ["post_date_gmt"]=>
      string(19) "2016-08-03 14:04:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(162) "NCT00679627 - A Randomized, Double-Blind, Placebo-controlled Trial of Long-term (2-year) Treatment of Galantamine in Mild to Moderately-Severe Alzheimer's Disease"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(155) "nct00679627-a-randomized-double-blind-placebo-controlled-trial-of-long-term-2-year-treatment-of-galantamine-in-mild-to-moderately-severe-alzheimers-disease"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2025-08-26 15:32:32"
      ["post_modified_gmt"]=>
      string(19) "2025-08-26 19:32:32"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(204) "https://dev-yoda.pantheonsite.io/clinical-trial/nct00679627-a-randomized-double-blind-placebo-controlled-trial-of-long-term-2-year-treatment-of-galantamine-in-mild-to-moderately-severe-alzheimers-disease/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
    [2]=>
    object(WP_Post)#5541 (24) {
      ["ID"]=>
      int(1848)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2019-12-12 13:04:00"
      ["post_date_gmt"]=>
      string(19) "2019-12-12 13:04:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(256) "NCT00574132 - A Phase 3, Multicenter, Randomized, Double-Blind, Placebo-Controlled, Parallel-Group, Efficacy and Safety Trial of Bapineuzumab (AAB-001, ELN115727) In Patients With Mild to Moderate Alzheimer's Disease Who Are Apolipoprotein E4 Non- Carriers"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(189) "nct00574132-a-phase-3-multicenter-randomized-double-blind-placebo-controlled-parallel-group-efficacy-and-safety-trial-of-bapineuzumab-aab-001-eln115727-in-patients-with-mild-to-moderate-alz"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2025-09-19 10:17:15"
      ["post_modified_gmt"]=>
      string(19) "2025-09-19 14:17:15"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(238) "https://dev-yoda.pantheonsite.io/clinical-trial/nct00574132-a-phase-3-multicenter-randomized-double-blind-placebo-controlled-parallel-group-efficacy-and-safety-trial-of-bapineuzumab-aab-001-eln115727-in-patients-with-mild-to-moderate-alz/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
    [3]=>
    object(WP_Post)#5542 (24) {
      ["ID"]=>
      int(1847)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2019-12-12 12:58:00"
      ["post_date_gmt"]=>
      string(19) "2019-12-12 12:58:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(251) "NCT00575055 - A Phase 3, Multicenter, Randomized, Double-Blind, Placebo-Controlled, Parallel-Group, Efficacy and Safety Trial of Bapineuzumab (AAB-001, ELN115727) In Patients With Mild to Moderate Alzheimer's Disease Who Are Apolipoprotein E4 Carriers"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(189) "nct00575055-a-phase-3-multicenter-randomized-double-blind-placebo-controlled-parallel-group-efficacy-and-safety-trial-of-bapineuzumab-aab-001-eln115727-in-patients-with-mild-to-moderate-alz"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2025-08-26 15:24:32"
      ["post_modified_gmt"]=>
      string(19) "2025-08-26 19:24:32"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(238) "https://dev-yoda.pantheonsite.io/clinical-trial/nct00575055-a-phase-3-multicenter-randomized-double-blind-placebo-controlled-parallel-group-efficacy-and-safety-trial-of-bapineuzumab-aab-001-eln115727-in-patients-with-mild-to-moderate-alz/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
    [4]=>
    object(WP_Post)#5545 (24) {
      ["ID"]=>
      int(1521)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2016-06-08 16:31:00"
      ["post_date_gmt"]=>
      string(19) "2016-06-08 16:31:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(155) "NCT00253214 - Placebo-Controlled Evaluation of Galantamine in the Treatment of Alzheimer's Disease: Safety and Efficacy of a Controlled-Release Formulation"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(151) "nct00253214-placebo-controlled-evaluation-of-galantamine-in-the-treatment-of-alzheimers-disease-safety-and-efficacy-of-a-controlled-release-formulation"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2025-08-26 15:30:52"
      ["post_modified_gmt"]=>
      string(19) "2025-08-26 19:30:52"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(200) "https://dev-yoda.pantheonsite.io/clinical-trial/nct00253214-placebo-controlled-evaluation-of-galantamine-in-the-treatment-of-alzheimers-disease-safety-and-efficacy-of-a-controlled-release-formulation/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
    [5]=>
    object(WP_Post)#5546 (24) {
      ["ID"]=>
      int(1327)
      ["post_author"]=>
      string(4) "1363"
      ["post_date"]=>
      string(19) "2015-01-30 12:12:00"
      ["post_date_gmt"]=>
      string(19) "2015-01-30 12:12:00"
      ["post_content"]=>
      string(0) ""
      ["post_title"]=>
      string(102) "NCT00253201 - Efficacy, Tolerability and Safety of Galantamine in the Treatment of Alzheimer's Disease"
      ["post_excerpt"]=>
      string(0) ""
      ["post_status"]=>
      string(7) "publish"
      ["comment_status"]=>
      string(6) "closed"
      ["ping_status"]=>
      string(6) "closed"
      ["post_password"]=>
      string(0) ""
      ["post_name"]=>
      string(98) "nct00253201-efficacy-tolerability-and-safety-of-galantamine-in-the-treatment-of-alzheimers-disease"
      ["to_ping"]=>
      string(0) ""
      ["pinged"]=>
      string(0) ""
      ["post_modified"]=>
      string(19) "2025-11-25 12:50:54"
      ["post_modified_gmt"]=>
      string(19) "2025-11-25 17:50:54"
      ["post_content_filtered"]=>
      string(0) ""
      ["post_parent"]=>
      int(0)
      ["guid"]=>
      string(147) "https://dev-yoda.pantheonsite.io/clinical-trial/nct00253201-efficacy-tolerability-and-safety-of-galantamine-in-the-treatment-of-alzheimers-disease/"
      ["menu_order"]=>
      int(0)
      ["post_type"]=>
      string(14) "clinical_trial"
      ["post_mime_type"]=>
      string(0) ""
      ["comment_count"]=>
      string(1) "0"
      ["filter"]=>
      string(3) "raw"
    }
  }
  ["project_title"]=>
  string(119) "Evaluating and Benchmarking the Use of Generated-Clinical Data for Alzheimer’s Disease Trials: a Retrospective Study."
  ["project_narrative_summary"]=>
  string(879) "Generated data are emerging as a way to enrich existing real data by inferring missing observations. They have notably been applied to health data, completing patient trajectories or correcting population representativity.  However, their evaluation is often based solely on its similarity to real data samples without consideration of its clinical utility. Here, we aim to evaluate state-of-the-art methods in reproducing clinical trial outcomes.



Some data generation methods can be applied to any dataset, while others use disease-specific models leveraging expert knowledge. Promising models have been proposed specifically for Alzheimer's Disease, making it an interesting disease for evaluating both general and disease-specific methods.



This project is a collaboration between the University of Paris-Cité and French Health Innovation Agency."
  ["project_learn_source"]=>
  string(9) "colleague"
  ["principal_investigator"]=>
  array(7) {
    ["first_name"]=>
    string(10) "Stéphanie"
    ["last_name"]=>
    string(14) "Allassonnière"
    ["degree"]=>
    string(3) "PhD"
    ["primary_affiliation"]=>
    string(25) "University of Paris-Cité"
    ["email"]=>
    string(34) "stephanie.allassonniere@u-paris.fr"
    ["state_or_province"]=>
    string(5) "Paris"
    ["country"]=>
    string(6) "France"
  }
  ["project_key_personnel"]=>
  array(2) {
    [0]=>
    array(6) {
      ["p_pers_f_name"]=>
      string(5) "Rémi"
      ["p_pers_l_name"]=>
      string(8) "Trimbour"
      ["p_pers_degree"]=>
      string(3) "PhD"
      ["p_pers_pr_affil"]=>
      string(25) "University of Paris-Cité"
      ["p_pers_scop_id"]=>
      string(0) ""
      ["requires_data_access"]=>
      string(3) "yes"
    }
    [1]=>
    array(6) {
      ["p_pers_f_name"]=>
      string(7) "Camille"
      ["p_pers_l_name"]=>
      string(7) "Schurtz"
      ["p_pers_degree"]=>
      string(6) "PharmD"
      ["p_pers_pr_affil"]=>
      string(35) "French Agency for Health Innovation"
      ["p_pers_scop_id"]=>
      string(0) ""
      ["requires_data_access"]=>
      string(2) "no"
    }
  }
  ["project_ext_grants"]=>
  array(2) {
    ["value"]=>
    string(3) "yes"
    ["label"]=>
    string(65) "External grants or funds are being used to support this research."
  }
  ["project_funding_source"]=>
  string(59) "Prairie : ANR-23-IACL-0008 (French national research grant)"
  ["project_date_type"]=>
  string(18) "full_crs_supp_docs"
  ["property_scientific_abstract"]=>
  string(1665) "Background:

Clinical trials generate essential evidence but face challenges such as short evaluation windows, under-representation of key populations, recruitment difficulties, and strict data-access constraints. Generated data offer a way to augment real datasets, but their validity requires evaluation beyond similarity metrics.

Objective:

To evaluate state-of-the-art data generation methods for reproducing clinical trial outcomes in Alzheimer’s disease and to develop an open-source benchmarking pipeline.

Study Design:

Retrospective study using participant-level data augmented with generated data from completed Alzheimer’s trials. The analysis includes (1) Generation: Create generated participants using multiple data-generation methods, (2) Cohort construction: Build augmented cohorts where different percentages of generated participants supplement either both arms or only the control arm, matched to baseline characteristics, (3) Evaluation: Compare trial outcomes and assess generated data quality using similarity and diversity metrics.

Participants:

Individuals with mild to severe Alzheimer’s disease in completed clinical trials.

Outcome Measures:

Primary: Accuracy of reproducing clinical outcome using generated+real vs. real-only datasets; Secondary: Evaluation of data utility and quality.

Statistical Analysis: Compare Absolute Risk Difference (ARD) and other endpoints between real and augmented cohorts; assess distribution similarity (e.g., Wasserstein distance) and diversity using metrics (e.g., Nearest Neighbor Distance Ratio and Distance to Closest Record."
  ["project_brief_bg"]=>
  string(3253) "Artificial data generation methods have rapidly advanced, enabling the creation of artificial patient records that preserve statistical and clinical properties of real patients (Nikolopoulos et al., 2024). These methods may (1) anonymize sensitive data, (2) increase cohort size, (3) balance underrepresented populations, and (4) increase statistical power by filling rare or unobserved clinical trajectories. Their adoption in health research is accelerating, especially in clinical trials where recruiting patients can be challenging (Briel et al., 2021).

Data generation methods can be categorized by their structure and generative process. Some models use real data to learn their statistical distribution (e.g. VAE, GAN) while others propose mathematical equations or mechanical models of the disease to sample new data points. Our team previously developed methods specifically trained with, and for, Alzheimer’s disease progression modelling (Koval et al., 2021; Chadebec et al., 2023). These models are representative of both categories, first by a differential equation system describing disease progression, then a variational auto-encoder (VAE) discriminating brain MRIs from Alzheimer’s disease patients and healthy patients.

However, most evaluations of generated data focus on statistical similarity to real records rather than assessing whether trial conclusions remain unchanged when generated participants are included. Without evaluation at the level of clinical endpoints, decision makers and sponsors remain cautious about generated clinical data (Allassonnière et al., 2024).

This project addresses this gap by benchmarking generated-data methods on their ability to reproduce trial outcomes, treatment effects, and risk differences in Alzheimer’s trials. The results could inform guidelines for when generated data can be trusted, how much can safely be used, and which algorithms are most reliable. This research project aims to help clinical researchers responsibly integrate generated data into future studies in three ways. (1) An open-source software to simplify the use of several data generation methodologies, (2) reproducible metrics to evaluate the quality of the generated data, and (3) methodological recommendations on the use of these data in clinical trials.



----------

Nikolopoulos, A. & Karalis, V. Implementation of a Generative AI Algorithm for Virtually Increasing the Sample Size of Clinical Studies. Appl. Sci. 14, 4570 (2024).

Briel, M. et al.Exploring reasons for recruitment failure in clinical trials: a qualitative study with clinical trial stakeholders in Switzerland, Germany, and Canada. Trials 22, 844 (2021).

Koval, I. et al. AD Course Map charts Alzheimer’s disease progression. Sci. Rep.11, 8020 (2021).

Chadebec, C., Thibeau-Sutre, E., Burgos, N. & Allassonnière, S. Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder. IEEE Trans. Pattern Anal. Mach. Intell. 45, 2879–2896 (2023).

Allassonnière, S. & Fraysse, J.-L. Données de santé artficielles : analyse et pistes de réflexion. https://static.botdesign.net/docs/Livre_blanc.pdf

"
  ["project_specific_aims"]=>
  string(546) "1. Benchmark multiple data generation algorithms on Alzheimer’s clinical trial data using clinically meaningful metrics.

2. Define a comprehensive scoring system to evaluate the utility, similarity, diversity, and reliability of generated data.

3. Establish evidence-based guidelines describing when and how generated data can safely enrich real clinical trial analyses.

4. Provide an open-source pipeline (Python/R) generating and evaluating generated clinical data, enabling reproducibility and future method development."
  ["project_study_design"]=>
  array(2) {
    ["value"]=>
    string(8) "meth_res"
    ["label"]=>
    string(23) "Methodological research"
  }
  ["project_purposes"]=>
  array(3) {
    [0]=>
    array(2) {
      ["value"]=>
      string(37) "develop_or_refine_statistical_methods"
      ["label"]=>
      string(37) "Develop or refine statistical methods"
    }
    [1]=>
    array(2) {
      ["value"]=>
      string(34) "research_on_clinical_trial_methods"
      ["label"]=>
      string(34) "Research on clinical trial methods"
    }
    [2]=>
    array(2) {
      ["value"]=>
      string(50) "research_on_clinical_prediction_or_risk_prediction"
      ["label"]=>
      string(50) "Research on clinical prediction or risk prediction"
    }
  }
  ["project_research_methods"]=>
  string(408) "We are going to replicate the studies by applying the same statistical analysis plan; no patients will be excluded on other criteria.



In addition, we will add virtual patients generated using AI-based augmentation.

When no control arm is available, the mechanistic models should also be tested on generating virtual control patients that follow the trajectory of the general population."
  ["project_main_outcome_measure"]=>
  string(986) "Our objective is to rerun each clinical trial according to its original statistical analysis plan and assess how closely the results obtained from real-only datasets are reproduced in different scenarios. We will consider both scenarios where whole cohorts are augmented and when only a part of the control arms is supplemented with generated data.



Primary outcome:

Ability of cohorts enriched with generated data to reproduce the original clinical trial conclusions, measured primarily by the difference in treatment effect estimates (e.g., Absolute Risk Difference) between real-only vs. generated+real datasets.

Secondary outcomes:

 • Distribution similarity metrics (e.g., Wasserstein distance)

 • Privacy and diversity metrics (Nearest Neighbor Distance Ratio, Distance to Closest Record) • Impact on subgroup analysis (mild vs. moderate vs. severe AD)

• Change in statistical power when generated samples are added

"
  ["project_main_predictor_indep"]=>
  string(648) "Our research will produce: generated clinical data files, a methodology for quantitative evaluation of their utility and quality. Since we focus on reproducing the clinical trial conclusions, the main independent variables of this study will be the generation method and the ratio of generated to real data.

We also aim to develop a framework characterizing a utility limit, defined as the maximum proportion of generated samples that can be added while preserving the utility of the dataset and maintaining concordance with the original clinical trial conclusions. This threshold may depend on the variability and complexity of the dataset."
  ["project_other_variables_interest"]=>
  string(1032) "Many demographic and outcome variables are necessary to evaluate the potential of data generation methods to enrich clinical data without introducing bias.

The following variables will be used when available:



Demographic Variables:

●	Age: Continuous variable (in years or months).

●	Sex: Categorical variable (Male/Female/Other).

●	Education Level: Ordinal variable (No formal education, Primary, Secondary, Higher).



Disease-related and Variable

●	Cognitive Measures useful to evaluate performances:

●	Baseline Cognitive Scores: Continuous variables (e.g., MMSE, ADAS-Cog, CDR-SB)

●	Disease Severity: Ordinal variable (e.g., Mild, Moderate, Severe based on clinical diagnosis).

●	Comorbidities: Binary variables (e.g., Diabetes: Yes/No, Hypertension: Yes/No).



Treatment Variables:

●	Medication Type: Categorical variable.

●	Treatment Duration: Continuous variable (measured in days or weeks)

"
  ["project_stat_analysis_plan"]=>
  string(3613) "To build generated patient data, we will use different methods to generate high-quality health data. We employ notably: (1) Variational Autoencoders (VAEs), (2) Graphical Adversarial Neural networks (GANs), (3) Mixed effect generative models, (4) Mechanistic models.

Each method will be tested within two frameworks: (1) augmenting both trial arms, and (2) augmenting only the control arm.

 1. VAE: Incorporates generative hierarchical models, enhancing flexibility in data sampling via latent variables with parametric distributions. It includes a model already trained to classify AD patients and healthy MRI (Chadebec et al., 2024).

2. GAN: Opposes two models, one generator of data and a classifier, which are trained together until the classifier cannot differentiate true and generated data.

3. Mixed-effect model: Allows for generating new patients after defining parameter values from populations. It includes the model already proposed by our team: http://www.digital-brain.org/ (Koval et al., 2021).

 4. Mechanistic model: Models the disease progression under drug effect. It includes the models proposed by the Critical Path Institute (https://c-path.org/program/critical-path-for-alzheimers-disease/) - access already granted for this project.

The statistical analysis plan will rely on the following two criteria: these criteria may be enriched along the study to better describe the effectiveness of the methods in the specific context of clinical trials.



1.	Fidelity

Fidelity will be evaluated by several statistical scores:  Mean (μ) of each variable of the patient vector, Standard Deviation (σ) for all variables independently, Skewness, Kurtosis, Frobenius distance between Covariance Matrices, Conditional Means, and covariances. For each statistical score, item-level thresholds are: 

 - First and second-order moments (mean and standard deviation): maximum deviation of 5% using Wasserstein Distance

 - Conditional First and second-order moments (mean and standard deviation): maximum deviation of 5%

In addition to these measures, an additional  test will be conducted:

 - Kolmogorov-Smirnov Test (KS Test), under the null hypothesis that the real and generated data have different distributions

 Global representativeness score acceptance thresholds are: 

 - Very Good Representativeness: Tolerance below 5% for all metrics and statistical tests.

 - Good Representativeness: Tolerance between 5% and 10% for all metrics and

 statistical tests.

 - Poor Representativeness: Tolerance above 10% for all metrics and statistical tests



2. Utility

Utility will be evaluated through clinical trial reproduction, by reproducing the same statistical endpoint analysis (e.g., ANCOVA, mixed-model repeated measures, Cox models, or ARD) that were presented in the conclusion of the clinical trials. The difference between with and without generated data enrichment will be quantified. Bootstrapping of both the generated and real data should be used to evaluate the stability of the results.



Additionally, several visualization tools (hierarchical clustering, heatmap, and PCA) will be used to communicate the evaluations graphically.



-------

Koval, I. et al. AD Course Map charts Alzheimer’s disease progression. Sci. Rep.11, 8020 (2021).

Chadebec, C., Thibeau-Sutre, E., Burgos, N. & Allassonnière, S. Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational"
  ["project_software_used"]=>
  array(4) {
    [0]=>
    array(2) {
      ["value"]=>
      string(6) "python"
      ["label"]=>
      string(6) "Python"
    }
    [1]=>
    array(2) {
      ["value"]=>
      string(1) "r"
      ["label"]=>
      string(1) "R"
    }
    [2]=>
    array(2) {
      ["value"]=>
      string(7) "rstudio"
      ["label"]=>
      string(7) "RStudio"
    }
    [3]=>
    array(2) {
      ["value"]=>
      string(11) "open_office"
      ["label"]=>
      string(11) "Open Office"
    }
  }
  ["project_timeline"]=>
  string(908) "●	November 2025: Clinical trials selection

●	December 2025: Selection and implementation of data generation methods to include

●	January 2026: Application of the methods to the YODA Alzheimer’s clinical trials

●	April 2026: First draft of the results on clinical trial outcome predictions. Definition of additional metrics to evaluate generated data quality (e.g., utility, balance between population categories, observed biases)

●	May 2026: Focus on underrepresented populations, evaluating method capacity in correcting clinical trial selection biases. 

●	June 2026: Implementation of the different methods and metrics to generate and evaluate the quality of generated clinical data in an open-source Python and/or R pipeline.

●	November 2026: Manuscript drafted and submitted

●	December 2026: Results reported back to YODA Project

"
  ["project_dissemination_plan"]=>
  string(701) "Results will be disseminated through one peer-reviewed journal manuscript and one open-source software release. Potential journals include Journal of Open-Source Software, Clinical Trials, Journal of Biomedical Informatics, or Artificial Intelligence in Medicine. The open-source benchmarking pipeline will be publicly released on GitHub, enabling other researchers as well as sponsors to evaluate generated clinical data methods on real trials. 

This project aims to provide guidelines supporting both researchers and regulators. Findings will thus also be presented at relevant conferences or workshops organised by international or national regulatory agencies on medical product validation."
  ["project_bibliography"]=>
  string(1536) "
 

Papadopoulos, D. & Karalis, V. D. Variational Autoencoders for Data Augmentation in Clinical Studies. Appl. Sci. 13, 8793 (2023).
Nikolopoulos, A. & Karalis, V. Implementation of a Generative AI Algorithm for Virtually Increasing the Sample Size of Clinical Studies. Appl. Sci. 14, 4570 (2024).
Briel, M. et al.Exploring reasons for recruitment failure in clinical trials: a qualitative study with clinical trial stakeholders in Switzerland, Germany, and Canada. Trials 22, 844 (2021).
Koval, I. et al. AD Course Map charts Alzheimer’s disease progression. Sci. Rep.11, 8020 (2021).
Chadebec, C., Thibeau-Sutre, E., Burgos, N. & Allassonnière, S. Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder. IEEE Trans. Pattern Anal. Mach. Intell. 45, 2879–2896 (2023).
Fernandes, A., Porcher, R., Tran, V.-T. & Petit, F. Evaluating virtual-control-augmented trials for reproducing treatment effect from original RCTs. Preprint at https://doi.org/10.48550/arXiv.2507.16048 (2025).
 Allassonnière, S. & Fraysse, J.-L. Données de santé artficielles : analyse et pistes de réflexion. https://static.botdesign.net/docs/Livre_blanc.pdf

"
  ["project_suppl_material"]=>
  bool(false)
  ["project_coi"]=>
  array(3) {
    [0]=>
    array(1) {
      ["file_coi"]=>
      array(21) {
        ["ID"]=>
        int(18388)
        ["id"]=>
        int(18388)
        ["title"]=>
        string(21) "allassonniere_coi.pdf"
        ["filename"]=>
        string(21) "allassonniere_coi.pdf"
        ["filesize"]=>
        int(20322)
        ["url"]=>
        string(70) "https://yoda.yale.edu/wp-content/uploads/2025/11/allassonniere_coi.pdf"
        ["link"]=>
        string(67) "https://yoda.yale.edu/data-request/2025-0796/allassonniere_coi-pdf/"
        ["alt"]=>
        string(0) ""
        ["author"]=>
        string(4) "2247"
        ["description"]=>
        string(0) ""
        ["caption"]=>
        string(0) ""
        ["name"]=>
        string(21) "allassonniere_coi-pdf"
        ["status"]=>
        string(7) "inherit"
        ["uploaded_to"]=>
        int(18260)
        ["date"]=>
        string(19) "2025-11-28 10:43:45"
        ["modified"]=>
        string(19) "2025-11-28 10:43:50"
        ["menu_order"]=>
        int(0)
        ["mime_type"]=>
        string(15) "application/pdf"
        ["type"]=>
        string(11) "application"
        ["subtype"]=>
        string(3) "pdf"
        ["icon"]=>
        string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
      }
    }
    [1]=>
    array(1) {
      ["file_coi"]=>
      array(21) {
        ["ID"]=>
        int(18389)
        ["id"]=>
        int(18389)
        ["title"]=>
        string(16) "Trimbour_coi.pdf"
        ["filename"]=>
        string(16) "Trimbour_coi.pdf"
        ["filesize"]=>
        int(17839)
        ["url"]=>
        string(65) "https://yoda.yale.edu/wp-content/uploads/2025/11/Trimbour_coi.pdf"
        ["link"]=>
        string(62) "https://yoda.yale.edu/data-request/2025-0796/trimbour_coi-pdf/"
        ["alt"]=>
        string(0) ""
        ["author"]=>
        string(4) "2247"
        ["description"]=>
        string(0) ""
        ["caption"]=>
        string(0) ""
        ["name"]=>
        string(16) "trimbour_coi-pdf"
        ["status"]=>
        string(7) "inherit"
        ["uploaded_to"]=>
        int(18260)
        ["date"]=>
        string(19) "2025-11-28 10:43:47"
        ["modified"]=>
        string(19) "2025-11-28 10:43:50"
        ["menu_order"]=>
        int(0)
        ["mime_type"]=>
        string(15) "application/pdf"
        ["type"]=>
        string(11) "application"
        ["subtype"]=>
        string(3) "pdf"
        ["icon"]=>
        string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
      }
    }
    [2]=>
    array(1) {
      ["file_coi"]=>
      array(21) {
        ["ID"]=>
        int(18390)
        ["id"]=>
        int(18390)
        ["title"]=>
        string(17) "DoI-C-Schurtz.pdf"
        ["filename"]=>
        string(17) "DoI-C-Schurtz.pdf"
        ["filesize"]=>
        int(18331)
        ["url"]=>
        string(66) "https://yoda.yale.edu/wp-content/uploads/2025/11/DoI-C-Schurtz.pdf"
        ["link"]=>
        string(63) "https://yoda.yale.edu/data-request/2025-0796/doi-c-schurtz-pdf/"
        ["alt"]=>
        string(0) ""
        ["author"]=>
        string(4) "2247"
        ["description"]=>
        string(0) ""
        ["caption"]=>
        string(0) ""
        ["name"]=>
        string(17) "doi-c-schurtz-pdf"
        ["status"]=>
        string(7) "inherit"
        ["uploaded_to"]=>
        int(18260)
        ["date"]=>
        string(19) "2025-11-28 10:43:48"
        ["modified"]=>
        string(19) "2025-11-28 10:43:50"
        ["menu_order"]=>
        int(0)
        ["mime_type"]=>
        string(15) "application/pdf"
        ["type"]=>
        string(11) "application"
        ["subtype"]=>
        string(3) "pdf"
        ["icon"]=>
        string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
      }
    }
  }
  ["data_use_agreement_training"]=>
  bool(true)
  ["human_research_protection_training"]=>
  bool(true)
  ["certification"]=>
  bool(true)
  ["search_order"]=>
  string(1) "0"
  ["project_send_email_updates"]=>
  bool(false)
  ["project_publ_available"]=>
  bool(true)
  ["project_year_access"]=>
  string(4) "2026"
  ["project_rep_publ"]=>
  bool(false)
  ["project_assoc_data"]=>
  array(0) {
  }
  ["project_due_dil_assessment"]=>
  array(21) {
    ["ID"]=>
    int(18656)
    ["id"]=>
    int(18656)
    ["title"]=>
    string(47) "YODA Project Due Diligence Assessment 2025-0796"
    ["filename"]=>
    string(51) "YODA-Project-Due-Diligence-Assessment-2025-0796.pdf"
    ["filesize"]=>
    int(130771)
    ["url"]=>
    string(100) "https://yoda.yale.edu/wp-content/uploads/2025/11/YODA-Project-Due-Diligence-Assessment-2025-0796.pdf"
    ["link"]=>
    string(93) "https://yoda.yale.edu/data-request/2025-0796/yoda-project-due-diligence-assessment-2025-0796/"
    ["alt"]=>
    string(0) ""
    ["author"]=>
    string(4) "1885"
    ["description"]=>
    string(0) ""
    ["caption"]=>
    string(0) ""
    ["name"]=>
    string(47) "yoda-project-due-diligence-assessment-2025-0796"
    ["status"]=>
    string(7) "inherit"
    ["uploaded_to"]=>
    int(18260)
    ["date"]=>
    string(19) "2026-01-26 13:34:52"
    ["modified"]=>
    string(19) "2026-01-26 13:34:52"
    ["menu_order"]=>
    int(0)
    ["mime_type"]=>
    string(15) "application/pdf"
    ["type"]=>
    string(11) "application"
    ["subtype"]=>
    string(3) "pdf"
    ["icon"]=>
    string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
  }
  ["project_title_link"]=>
  array(21) {
    ["ID"]=>
    int(18657)
    ["id"]=>
    int(18657)
    ["title"]=>
    string(46) "YODA Project Protocol - 2025-0796 - 2025-12-10"
    ["filename"]=>
    string(46) "YODA-Project-Protocol-2025-0796-2025-12-10.pdf"
    ["filesize"]=>
    int(187156)
    ["url"]=>
    string(95) "https://yoda.yale.edu/wp-content/uploads/2025/11/YODA-Project-Protocol-2025-0796-2025-12-10.pdf"
    ["link"]=>
    string(88) "https://yoda.yale.edu/data-request/2025-0796/yoda-project-protocol-2025-0796-2025-12-10/"
    ["alt"]=>
    string(0) ""
    ["author"]=>
    string(4) "1885"
    ["description"]=>
    string(0) ""
    ["caption"]=>
    string(0) ""
    ["name"]=>
    string(42) "yoda-project-protocol-2025-0796-2025-12-10"
    ["status"]=>
    string(7) "inherit"
    ["uploaded_to"]=>
    int(18260)
    ["date"]=>
    string(19) "2026-01-26 13:35:20"
    ["modified"]=>
    string(19) "2026-01-26 13:35:20"
    ["menu_order"]=>
    int(0)
    ["mime_type"]=>
    string(15) "application/pdf"
    ["type"]=>
    string(11) "application"
    ["subtype"]=>
    string(3) "pdf"
    ["icon"]=>
    string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
  }
  ["project_review_link"]=>
  array(21) {
    ["ID"]=>
    int(18658)
    ["id"]=>
    int(18658)
    ["title"]=>
    string(36) "YODA Project Review - 2025-0796_site"
    ["filename"]=>
    string(38) "YODA-Project-Review-2025-0796_site.pdf"
    ["filesize"]=>
    int(1331883)
    ["url"]=>
    string(87) "https://yoda.yale.edu/wp-content/uploads/2025/11/YODA-Project-Review-2025-0796_site.pdf"
    ["link"]=>
    string(80) "https://yoda.yale.edu/data-request/2025-0796/yoda-project-review-2025-0796_site/"
    ["alt"]=>
    string(0) ""
    ["author"]=>
    string(4) "1885"
    ["description"]=>
    string(0) ""
    ["caption"]=>
    string(0) ""
    ["name"]=>
    string(34) "yoda-project-review-2025-0796_site"
    ["status"]=>
    string(7) "inherit"
    ["uploaded_to"]=>
    int(18260)
    ["date"]=>
    string(19) "2026-01-26 13:35:44"
    ["modified"]=>
    string(19) "2026-01-26 13:35:44"
    ["menu_order"]=>
    int(0)
    ["mime_type"]=>
    string(15) "application/pdf"
    ["type"]=>
    string(11) "application"
    ["subtype"]=>
    string(3) "pdf"
    ["icon"]=>
    string(62) "https://yoda.yale.edu/wp/wp-includes/images/media/document.png"
  }
  ["project_highlight_button"]=>
  string(0) ""
  ["request_overridden_res"]=>
  string(1) "3"
  ["request_data_partner"]=>
  string(15) "johnson-johnson"
}
data partner
array(1) {
  [0]=>
  string(15) "johnson-johnson"
}


pi country
array(0) {
}


pi affil
array(0) {
}


products
array(2) {
  [0]=>
  string(8) "razadyne"
  [1]=>
  string(3) "n-a"
}


num of trials
array(1) {
  [0]=>
  string(1) "6"
}


res
array(1) {
  [0]=>
  string(1) "3"
}