Personalized Exercise Recommendations for Chronic Pelvic Pain Using Reinforcement Learning

Purpose

WorkoutCPP is a pilot study evaluating the feasibility of a personalized exercise recommendation system for individuals with chronic pelvic pain disorders (CPPDs). The study uses reinforcement learning (RL), a type of artificial intelligence that adapts recommendations over time based on each participant's reported pain levels, symptom burden, and exercise compliance. Participants receive daily exercise recommendations that alternate between standard, non-personalized guidance and personalized, RL-generated recommendations across four 2-week phases, allowing within-person comparison of outcomes under each condition. The primary hypothesis is that an RL-based adaptive recommendation system is feasible to deliver in a CPPD population.

Conditions

  • Pelvic Pain
  • Endometriosis
  • Chronic Pelvic Pain

Eligibility

Eligible Ages
Between 18 Years and 55 Years
Eligible Sex
Female
Accepts Healthy Volunteers
No

Inclusion Criteria

  • Self-reported CPPD (e.g., endometriosis, adenomyosis, fibroids, etc.) based on clinician diagnosis - Aged 18-55 years. - Ownership of an iOS or Android smartphone. - Willingness to self-track daily symptoms, exercise activities, and self-management behaviors using a smartphone research app. - Willingness to wear an activity tracker for the study duration. - Willingness to follow exercise recommendations from a smartphone research app, provided no adverse symptoms occur. - Ability to read and write in English sufficient to understand study materials and communications. - At least intermittently physically active (e.g., ≥30 minutes of walking twice per week).

Exclusion Criteria

  • Absolute contraindications to PA (e.g., recent myocardial infarction, complete heart block, acute congestive heart failure, unstable angina, or uncontrolled severe hypertension, BP ≥180/110 mm Hg). - More than two "Yes" responses on the Physical Activity Readiness Questionnaire (PAR-Q) (16) without physician clearance. - Major life events expected during the next 10 weeks (e.g., pregnancy, planned surgery, or extended travel likely to interfere with participation). - Current or planned pregnancy within the next 6 months. - Having given birth in the past 6 months or currently nursing. - Inability to wear an activity tracker or use the app for the study duration. - Complete inactivity (i.e., <60 minutes of moderate-intensity PA per week).

Study Design

Phase
N/A
Study Type
Interventional
Allocation
Randomized
Intervention Model
Crossover Assignment
Intervention Model Description
N-of-1 randomized crossover design in which each participant serves as their own control, randomized 1:1 to one of two intervention sequences (ABAB or BABA), alternating between RL-generated personalized exercise recommendations (active arm) and standard generic recommendations (control arm) across multiple 2-week blocks within the 9-week study period. First week is treated as a baseline week where participants get accommodated to the study App and procedures, as well as RL agent warm-up.
Primary Purpose
Other
Masking
Single (Participant)
Masking Description
Participants are blinded to their assigned study phase (active control: generic exercise recommendations vs. experimental: RL-generated adaptive recommendations) and are not informed of the phase sequence or their current assignment at any point during the study. However, participants may be able to infer their assigned phase over time based on the nature of the recommendations received. Study investigators and the data analysis team are not blinded to phase assignment.

Arm Groups

ArmDescriptionAssigned Intervention
Experimental
RL-based personalized phase
Participants will receive RL-generated personalized exercise recommendations, which are generated using the list from the initial participant intake form indicating their capacity and resources for carrying out various modalities and intensities of physical activity. The RL agent learns from the participant feedback to update the update the subsequent recommendations.
  • Behavioral: Reinforcement Learning (RL)-Based Personalized Exercise Recommendations
    Daily exercise recommendations (using type, intensity, and duration) are generated by a contextual bandit reinforcement learning agent, based on the implementation described in Meier et al. 2023. Recommendations are personalized using each participant's initially generated list of exercises based on their physical ability and resources available, as well as contextual daily factors including pain symptoms, prior exercise compliance, and their feedback to the previous exercise recommendation.
Active Comparator
Standard (Generic) Exercise Arm
Participants will receive standardized, non-personalized exercise recommendations based on the U.S. Physical Activity Guidelines, in 2-week blocks. This comparison will serve as the "active control" arm to which the experimental RL arm will be compared. This type of control condition was selected to provide a more rigorous test of the experimental condition.
  • Behavioral: Generic Exercise Recommendation
    Participants receive exercise recommendations from a standardized, set list of exercise recommendations that are based on USDHHS physical activity guidelines (Piercy et al., 2020). Recommendations are not personalized based on participant contextual information and do not adapt over the course of the study.

Recruiting Locations

Icahn School of Medicine at Mount Sinai
New York, New York 10029
Contact:
Ipek Ensari, PhD
631-565-1829
ipek.ensari@mssm.edu

More Details

Status
Recruiting
Sponsor
Icahn School of Medicine at Mount Sinai

Study Contact

Ipek Ensari, PhD
631-565-1829
ipek.ensari@mssm.edu

Detailed Description

Chronic pelvic pain disorders (CPPDs) are associated with high symptom burden and reduced quality of life. Physical activity (PA) and exercise have emerged as a promising non-pharmacological approach for symptom management. However, optimal exercise type, intensity, and timing for pain management vary substantially across individuals, supporting the need for personalized adaptive approaches (Ensari et al., 2022, Krasny-Pacini et al., 2017). This study will enroll participants will a CPPD diagnosis into a remote, 9-week study to evaluate the feasibility of RL-based personalized exercise to non-personalized, standard recommendations. Enrollment is rolling, with participants entering the study on a continuous basis. Each participant's start date, and their 9-week intervention period, is determined by their baseline interview date. A baseline interview upon enrollment is scheduled with an exercise physiologist to review the participant's initial exercise list and provide exercise safety information, as well as overview use of the study App. Participants can choose to stay in the study for 2 additional weeks to make up any weeks with inadequate adherence. Study outcomes are measured daily over the course of the intervention period. Daily App-based tracking items assess pain and other symptoms, exercise behavior, perceived effect and feedback to the recommendation, menstrual status, and recommendation compliance. Fitbit trackers simultaneously track participants' objectively-estimated PA. A reinforcement learning (RL) agent implemented in Meier et al. 2023 as the middleware platform generates daily personalized exercise recommendations delivered via a research mobile phone application (Hirten et al., 2023, Meier et al., 2023). Participant-reported perceived effect of each exercise recommendation is used by the RL agent to calculate reward. Participants serve as their own controls, allowing for within-person comparison under the two conditions (Krasny-Pacini et al., 2017). Primary outcomes for the study include standard study feasibility metrics (e.g., adherence, retention). Secondary outcomes focus on RL agent performance and learning over time. Participant safety will be monitored throughout the study, in accordance with the institutional review board. This work was supported by the Digital Health Partnership (DHP), a collaboration between the Hasso Plattner Institute, Data4Life, the Windreich Department of Artificial Intelligence and Human Health, the Hasso Plattner Institute for Digital Health at Mount Sinai, and The Charles Bronfman Institute for Personalized Medicine at the Icahn School of Medicine at Mount Sinai.