MishenMed: Statistical Consulting for Clinical Trials

 Finding unusual events: Gaussian (Normal) Distribution: Anomaly detection algorithm: Developing and evaluating an anomaly detection system: Anomaly detection v.s. Supervised learning Anomaly detection is designed to identify novel positive instances that may not resemble any previously observed examples. In contrast, supervised learning uses known positive examples to determine whether future instances are similar to what has been seen before. When building an anomaly detection algorithm, choosing what features to use ? Non-gaussian features The typical development process involves training the model, then reviewing the anomalies in the cross-validation set that the algorithm fails to detect. By examining these missed examples, you can identify potential new features. For instance, if an anomaly exhibits unusually high or low values on a newly engineered feature, that feature can help the algorithm successfully flag similar anomalies.

In a group sequential design, interim analyses are planned to potentially stop a trial early either because the treatment appears promising (efficacy) or because it seems unlikely to show a benefit (futility). The key differences in how these analyses affect the type I error (α) are as follows: Futility Analysis Purpose: Futility analyses are used to stop a trial early when data suggest that it is unlikely to achieve a statistically significant benefit. The goal here is to avoid exposing patients to an ineffective treatment and to conserve resources. Alpha Impact: Since futility decisions do not lead to a claim of efficacy, they do not contribute to an inflation of the type I error rate. In other words, stopping a trial early for futility does not “risk” a false positive finding because no positive claim is being made. Non-Binding Nature: Often, futility boundaries are set as non-binding. This means that even if the futility criterion is met, investigators might choose to continue t...

 Company Core Data Sheet (CCDS) An evidence-based reference document that comprises company's global view of the profile of a specific product and represents clinical, nonclinical, scientific, and technical data, and other relevant information. The CCDS describes the full scope of benefit information based on adequate data and establishes the essential safety information (CCSI) for regulatory submission in all local or regional product labeling. The CCDS is a reference for the development of local labeling, as well as periodic safety reports. Company Core Safety Information (CCSI) For medicinal products, all relevant safety information contained in the company core data sheet prepared by the marketing authorization holder and which the marketing authorization holder requires to be listed in all countries where the company markets the product, except when the local Regulatory Authority specifically requires a modification (GVP Annex IV, ICH-E2C(R2) Guideline).  It is the refere...

  What is a Lymph Node? A lymph node is a small, bean-shaped organ that is part of the lymphatic system . Lymph nodes filter lymph fluid , trapping bacteria, viruses, cancer cells, and other harmful substances. They play a crucial role in immune response and cancer staging . 1. Function of Lymph Nodes ✔ Filter harmful substances from lymph fluid before it returns to the bloodstream. ✔ Store and activate immune cells (T cells, B cells, macrophages) to fight infections. ✔ Indicate disease presence (e.g., infection, inflammation, or cancer metastasis). 2. Location of Lymph Nodes in the Body There are hundreds of lymph nodes throughout the body, clustered in key regions: Lymph Node Group Location Clinical Importance Cervical Nodes Neck Swollen in throat infections, head & neck cancer Axillary Nodes Armpits Common site for breast cancer metastasis Inguinal Nodes Groin Filter lymph from lower limbs, genitals Mediastinal Nodes Chest Important in lung cancer staging Mesenteric No...

  What is Tumor Size Based On in Oncology Trials? In oncology clinical trials , tumor size is typically measured using radiological imaging techniques and is based on standardized criteria , such as RECIST (Response Evaluation Criteria in Solid Tumors) . The tumor size used in waterfall plots and treatment response assessments is derived from longest diameters or volumetric measurements of target lesions. 1. Measurement of Tumor Size Tumor size is usually assessed through radiologic imaging techniques , including: CT Scan (Computed Tomography) MRI (Magnetic Resonance Imaging) PET Scan (Positron Emission Tomography) Ultrasound (for specific tumors) These imaging methods help oncologists measure the tumor’s longest diameter and monitor changes over time. 2. RECIST Criteria for Tumor Size Assessment RECIST (Response Evaluation Criteria in Solid Tumors) is the most widely used standard for measuring tumor size and treatment response. It defines tumor size as: Sum of the longest d...

  Waterfall Plot in Oncology Trials:  What is a Waterfall Plot in Oncology? A waterfall plot is a bar chart commonly used in oncology clinical trials to visually represent tumor response for individual patients. Each bar represents one patient , showing how much their tumor size has changed from baseline after treatment. Key Features of a Waterfall Plot X-Axis (Patients) : Each bar represents an individual patient. Y-Axis (Tumor Size Change, % Reduction/Growth from Baseline) : Negative values (below zero) → Tumor shrinking (partial or complete response) . Positive values (above zero) → Tumor growth or progression . Bars reaching or exceeding -30% → Meet the criteria for Partial Response (PR) per RECIST (Response Evaluation Criteria in Solid Tumors) . Bars reaching or exceeding +20% → Indicate Progressive Disease (PD) . How to Interpret the Y-Axis? The y-axis represents the percentage change in tumor size from baseline. The percentage is calculated as: Percentage...

  Simon’s Two-Stage Design is commonly used in phase II clinical trials to evaluate whether a new treatment shows sufficient promise for further study while minimizing the number of patients exposed to ineffective treatments. The design involves: An initial stage (Stage 1) where a small number of patients are enrolled and assessed. If promising results are observed, the trial proceeds to Stage 2 with additional patients. If the treatment is ineffective in Stage 1 , the trial is stopped early. However, in real-world clinical trials, overruns (exceeding planned sample size) or underruns (fewer patients than planned) can occur due to logistical, operational, or recruitment issues . These deviations can impact the Type I error rate (α level) and the statistical integrity of the trial. 1. Impact of Overrun or Underrun on Alpha Control A. Overrun (More Patients Enrolled) Problem: If more patients are enrolled in Stage 1 , there is a higher chance of incorrectly rejecting the null ...

  Submitting data and documents to the U.S. Food and Drug Administration (FDA) is a crucial step in gaining regulatory approval for drugs, biologics, medical devices, and other healthcare products. The submission process ensures that products are safe, effective, and compliant with regulations before they reach the market. 1. Types of FDA Submissions A. Drug and Biologic Submissions Investigational New Drug (IND) Application Required before conducting clinical trials in humans. Includes preclinical data, manufacturing details, and study protocols . New Drug Application (NDA) Submitted for small-molecule drugs seeking marketing approval. Includes clinical trial results, safety and efficacy data, labeling, and manufacturing details . Biologics License Application (BLA) Required for biologic products (e.g., vaccines, gene therapies, monoclonal antibodies). Similar to an NDA but specific to biologics. 505(b)(2) NDA Submission Used for drugs with modifications to existing approved d...

  What is Causal Mediation Analysis? Causal mediation analysis is a statistical method used to determine how an independent variable (X) influences an outcome variable (Y) through an intermediate variable (M) , known as the mediator . This method helps researchers understand the mechanism behind causal effects. For example, if we want to study how exercise (X) affects heart health (Y) , mediation analysis can determine whether the effect is partially or fully explained by an intermediate factor like weight loss (M) . Key Components of Causal Mediation Analysis Exposure (X) : The independent variable (e.g., Exercise) Mediator (M) : The variable that transmits part of the effect (e.g., Weight Loss) Outcome (Y) : The dependent variable (e.g., Heart Health) The goal is to decompose the total effect of X  on Y  into: Direct Effect (DE) : The effect of X X  on Y Y   not passing through M . Indirect Effect (IE) / Mediated Effect : The portion of the effect tha...

  Both the Log-Rank Test and the Cox Proportional Hazards (Cox PH) Model are used in survival analysis to compare survival times between groups, but they have key differences in methodology, assumptions, and applications. 1. Log-Rank Test Purpose: Compares two or more survival curves (e.g., treatment vs. control) to determine if there is a statistically significant difference in survival times. Methodology: It is a non-parametric test based on comparing the observed vs. expected number of events (e.g., deaths, progression) at each time point. Uses a chi-square test statistic to determine significance. Key Assumptions: ✅ Proportional hazards assumption is NOT required . ✅ Works well when the hazard ratio (HR) is constant over time. ❌ Cannot adjust for covariates (e.g., age, sex, biomarkers). Example Interpretation: p < 0.05 : There is a significant difference between survival curves. p ≥ 0.05 : No significant difference detected. 2. Cox Proportional Hazards (Cox PH) Mod...