Effectiveness and Performance of an Optical Biopsy Technology for Esophageal Cancer in Brazil and the United States
- Conditions
- Suspected or Known Squamous Cell NeoplasiaPrior History of Squamous Cell Dysplasia and /or Neoplasia
- Interventions
- Device: Artificial Intelligence Mobile High-Resolution Microendoscope
- Registration Number
- NCT06435286
- Lead Sponsor
- Baylor College of Medicine
- Brief Summary
In a previous clinical trial in China and the United States (US), the investigators developed and validated a mobile, high-resolution microendoscope (mHRME) for screening and surveillance of esophageal squamous cell neoplasia (ESCN). The trial revealed higher specificity for qualitative (visual) interpretation by experts but not the novice and in the surveillance arm (100% vs. 19%, p \<0.05). In the screening arm, diagnostic yield (neoplastic biopsies/total biopsies) increased 3.6 times (8 to 29%); 16% of patients were correctly spared any biopsy, and 18% had a change in clinical plan. In a pilot study in Brazil, the investigators tested a software-assisted mHRME with deep-learning software algorithms to aid in the detection of neoplastic images and determine the performance, efficiency, and impact of the AI-mHRME when to Lugol's chromoendoscopy (LCE) alone and when using AI-mHRME with LCE. In this clinical trial, the investigators will build on the Brazil pilot trial data to optimize an artificial intelligence (AI) mHRME and evaluate its clinical impact and implementation potential in ethnically and socioeconomically diverse populations in the US and Brazil.
- Detailed Description
The investigators' hypothesis is that the artificial intelligence (AI) mobile, high-resolution microendoscope (mHRME) will increase the accuracy of Lugol's chromoendoscopy (LCE) in endoscopic cancer detection in low- and middle-income countries (LMICs) and high-income countries (HICs).
Objective 1: The investigators' first objective is to evaluate the diagnostic performance, efficiency, and impact of this automated optical biopsy device. In a single-arm study (n=200) of high-risk subjects undergoing LCE followed by AI-mHRME for ESCN screening in Brazil and the US, the investigators will evaluate the diagnostic performance and efficiency of this automated optical biopsy device.
The investigators' other hypotheses are that the AI-mHRME will:
1. increase the mHRME accuracy in novices and be non-inferior to experts,
2. increase user confidence among experts and novices, and
3. increase the LCE efficiency and impact byreducing biopsies and second procedures.
The investigators will compare the accuracy of the AI-mHRME software read to novice and expert clinicians' subjective reading to gold-standard histopathology by an expert gastrointestinal (GI) pathologist. For clinician confidence and clinical impact, they will determine the clinician's confidence level in the software diagnosis and the potential clinical impact of this diagnosis among novice and expert endoscopists using AI-mHRME. The clinician reads will be part of the mHRME procedure and treatment "plan" (biopsy vs. not biopsy vs. treat). Clinicians are not considered study subjects in objective 1. The clinical impact will be determined by the change in the clinician's decision in the treatment "plan" before and after the AI-mHRME read. For efficiency (biopsy saving and diagnostic yield), they will determine the number of patients spared any biopsy due to AI-mHRME. The investigators will compare the diagnostic yield of AI-mHRME and LCE vs. LCE alone (diagnostic yield = neoplastic biopsies/total number of biopsies obtained in biopsied patients).
Objective 2: This objective will have three study populations, with a total sample size of n=50 subjects. To determine barriers and facilitators to implementing AI-mHRME, the team will form Health Sector Stakeholder Advisory Boards (HS-SAB) in the US and Brazil as the first study population. The HS-SABs will include academic partners, primary care providers referring patients, doctors performing esophageal cancer screening, hospital administrators, and patient and caregiver representatives. The HS-SAB sample size will be 6-10 members in the US and Brazil each, a standard number of participants for research advisory boards. The team will collect feedback and input through focus group discussions (FGDs) at 6 time points across the project period per HS-SAB. FGD objectives will match the research stage: clinical trial planning (recruitment and retention plan refinement), data collection (stakeholders identification), result interpretation, and dissemination.
For the second study population, the team will conduct semi-structured individual interviews with implementers to assess barriers and facilitators to implementing AI-assisted cancer technologies (n=40). Interviews will be with patients and caregivers(n=10), GI clinicians (n=10), primary care physicians (n=10), and hospital and health leadership (n=10).
There will be surveys with endoscopists (n=40) at the participating sites to understand their thoughts on HRME.
Recruitment & Eligibility
- Status
- RECRUITING
- Sex
- All
- Target Recruitment
- 200
- Outpatients undergoing routine (standard of care) Lugol's chromoendoscopic screening for squamous cell neoplasia will be eligible for enrollment, including patients with a known history of head/neck squamous cell cancer; heavy smoking and alcohol, other dietary or geographic risk factors or prior dysplasia
- Patients >18 years old.
- Patients of any sex or gender.
- Patients who are willing and able to give informed consent.
- Allergy or prior reaction to the fluorescent contrast agent proflavine hemisulfate.
- Patients who are unable to give informed consent.
- Known advanced squamous cell carcinoma of the distal esophagus or dysplastic/suspected malignant esophageal lesion greater than or equal to 2 cm in size not amenable to endoscopic therapy.
- Patient unable to undergo routine endoscopy with biopsy:
- Women who are pregnant or breast feeding,
- Prothrombin time greater than 50% of control; PTT greater than 50 sec, or INR greater than 2.0,
- Inability to tolerate sedated upper endoscopy due to cardio-pulmonary instability or other significant medical issues.
Study & Design
- Study Type
- INTERVENTIONAL
- Study Design
- SINGLE_GROUP
- Arm && Interventions
Group Intervention Description Artificial Intelligence Mobile High Resolution Microendoscope (AI-mHRME) imaging Proflavine Hemisulfate All subjects will receive White Light Imaging (WLI) and Lugol's Chromoendoscopy (LCE), the current standard of care (SOC) procedure. Following LCE, all subjects will receive the artificial intelligence (AI) mobile high-resolution microendoscopy (mHRME) imaging with Proflavine Hemisulfate of any LCE abnormal and LCE normal areas (4:1 ratio). For both WLI and LCE, we will record the subjective clinician read (neoplastic, non-neoplastic), the confidence level in their diagnoses (high, low), and the action plan (biopsy vs. no biopsy vs. treat). With the AI-mHRME, we will image the same LCE abnormal and normal areas and record the software read, the clinician confidence level, and action plan. Finally, the imaged LCE abnormal areas will be biopsied or resected, and evaluated by a pathologist. Artificial Intelligence Mobile High Resolution Microendoscope (AI-mHRME) imaging Artificial Intelligence Mobile High-Resolution Microendoscope All subjects will receive White Light Imaging (WLI) and Lugol's Chromoendoscopy (LCE), the current standard of care (SOC) procedure. Following LCE, all subjects will receive the artificial intelligence (AI) mobile high-resolution microendoscopy (mHRME) imaging with Proflavine Hemisulfate of any LCE abnormal and LCE normal areas (4:1 ratio). For both WLI and LCE, we will record the subjective clinician read (neoplastic, non-neoplastic), the confidence level in their diagnoses (high, low), and the action plan (biopsy vs. no biopsy vs. treat). With the AI-mHRME, we will image the same LCE abnormal and normal areas and record the software read, the clinician confidence level, and action plan. Finally, the imaged LCE abnormal areas will be biopsied or resected, and evaluated by a pathologist.
- Primary Outcome Measures
Name Time Method Clinical Impact 18 months Change in clinical plan ('biopsy vs. no biopsy vs. treat') following AI-mHRME.
Clinician Confidence 18 months Confidence of expert and novice clinicians in clinically interpreting mHRME (pre- and post-use of AI-mHRME).
Performance Characteristics 18 months Sensitivity, specificity, positive and negative predictive values of AI-mHRME.
Procedure Efficiency 18 months Efficiency in the number of biopsies saved, procedures saved.
- Secondary Outcome Measures
Name Time Method
Trial Locations
- Locations (4)
Baylor St. Luke's Medical Center
🇺🇸Houston, Texas, United States
Ben Taub Hospital (Harris Health Systems)
🇺🇸Houston, Texas, United States
Hospital de Cancer de Barretos - Fundacao Pio XII
🇧🇷Barretos, São Paulo, Brazil
Instituto do Câncer do Estado de São Paulo
🇧🇷São Paulo, Brazil