Systematic Machine Learning Algorithm for Rapid Thrombosis Detection (DVT-SMART)

Evaluating a New Diagnostic Strategy for Suspected DVT Consisting of Point of Care D-dimer, AI-based Prediction Model and Compression Ultrasound

The goal of this clinical trial is to compare the use of a machine learning-based algorithm and point-of-care D-dimer to laboratory D-dimer and compression ultrasound to exclude deep vein thrombosis in the under extremities in patients referred to a medical department suspected of having deep vein thrombosis. The main aim is to answer are if a machine learning algorithm and point of care D-dimer can exclude deep vein thrombosis in more patients than clinical assessment and D-dimer alone.

Study Overview

• Status: Recruiting
• Conditions: Deep Vein Thrombosis
• Intervention / Treatment: Diagnostic test: POC D-dimer; Diagnostic test: POC ultrasound; Diagnostic test: Machine learning model

Detailed Description

All participants will follow the usual diagnostic algorithm used for patients with suspected DVT referred to Ostfold Hospital (all patients are examined by a physician, D-dimer is analyzed in all patients, ultrasound is performed by a radiologist in patients with positive D-dimer). In addition to usual care, POC D-dimer, POC ultrasound (performed by ED physicians), blood sampling for biobanking, and photographies of the under extremities will be performed. The machine learning model will be tested to see if the prediction is correct. In participants where ultrasound is performed, it will also be assessed whether the machine learning algorithm could have excluded the participant without the use of ultrasound. None of the additional procedures will have any impact on the patient diagnostics or treatment.

• Study Type: Interventional
• Enrollment (Estimated): 1000
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Waleed Ghanima, Professor; Phone Number: 0047 69860000; Email: waleed.ghanima@so-hf.no
• Study Contact Backup: Name: Hans Joakim Myklebust-Hansen, Medical Doctor; Phone Number: 0047 97501765; Email: hans.joakim.hansen@so-hf.no

Study Locations

• Norway
  • Sarpsborg, Norway, 1714
    • Recruiting
    • Østfold Hospital Trust
    • Contact: Hans J Myklebust-Hansen, M.D.; Phone Number: 004797501765; Email: hansjoakim_mh@outlook.com
    • Contact: Hans J Myklebust-Hansen, M.D.

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

• Patients referred to the ED due to suspicion of DVT
• Age ≥ 18 years
• Able to give informed consent

Exclusion Criteria:

• Ongoing use of anticoagulation for more than 72 hours
• Previous participation in the study
• Life expectancy of less than three months.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Diagnostic
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm

Intervention / Treatment

Experimental: All participants; All participants will be treated the same way.

Diagnostic test: POC D-dimer; POC D-dimer will be compared to laboratory D-dimer in hospital setting and used in a machine learning model; Diagnostic test: POC ultrasound; Point of care (POC) ultrasound performed by ED physicians compared to ultrasound performed by radiologist. POC ultrasound 3 point examination performed by ED physician will be compared with POC ultrasound full leg examination performed by ED physician.; Diagnostic test: Machine learning model; The DSS will be compared to the usual strategy. It will also be estimated how many participants where DVT could have been excluded without ultrasound.; Other Names: Decision support system

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Safety of the new strategy (POC D-dimer, ML-based prediction model, POC CUS by emergency physician)	Evaluate the safety of a new strategy consisting of POC D-dimer and an ML-based prediction model followed by CUS performed by emergency physicians by comparing the new strategy's safety with our standard care by measuring the proportion of patients in whom DVT is excluded according to the new strategy but was diagnosed with DVT by standard care or in whom DVT is diagnosed within the 90-day follow up.	From enrollment to the end of the primary assessment period (90 days)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Evaluate the efficiency of the new strategy	The proportion of patients in whom DVT can be ruled out by the ML-based prediction model with POC D-dimer compared to the efficiency of Wells score and laboratory D-dimer	From enrollment to the end of the primary assessment period (90 days)
Validate the safety and efficiency of the ML-based prediction model	Safety will be determined by the proportion of patients in whom DVT is excluded by the ML-model but diagnosed by standard care. Efficiency will be determined by the proportion of patients in whom DVT can be excluded by the ML-based model	From enrollment to the end of the primary assessment period (90 days)
Evaluate concordance between CUS performed by emergency physicians and radiologists.	Determine the proportion of false negative and false positive diagnosis of DVT in emergency physician-performed ultrasound compared with ultrasound performed by radiologists.	From time of enrollment until time of ultrasound examination performed by radiologist, assessed up to 48 hours.
Evaluate concordance between POC D-dimers in an ED setting and laboratory D-dimers.	Compare the two POC D-dimers with the STA-Liatest D-dimer and Siemens INNOVANCE by direct comparison of the true/false positive/negative results.	From enrollment to the completion of D-dimer analysis, assessed up to 24 hours.
Evaluate the hypothetical time to be completed for the novel strategy compared to the standard strategy.	Estimating the total management time defined as time from ED registration to ED discharge in patients evaluated according to the new strategy compared to standard care.	From time of enrollment until time of discharge from the emergency department either discharged from the hospital or hospitalized, assessed up to 24 hours.
Evaluate the safety of a limited ultrasound protocol (two-point and proximal) compared to full-leg CUS performed by emergency physicians and radiologists	Estimating the proportion of patients in whom DVT was ruled out by the limited ultrasound protocol but was diagnosed with DVT by the whole-leg ultrasound.	90 days after enrollment.

Collaborators and Investigators

• Sponsor: Ostfold Hospital Trust
• Collaborators: Sahlgrenska University Hospital

Study record dates

Study Major Dates

• Study Start (Actual): January 6, 2025
• Primary Completion (Estimated): January 5, 2027
• Study Completion (Estimated): January 5, 2029

Study Registration Dates

• First Submitted: December 4, 2024
• First Submitted That Met QC Criteria: February 18, 2025
• First Posted (Actual): March 25, 2025

Study Record Updates

• Last Update Posted (Actual): March 25, 2025
• Last Update Submitted That Met QC Criteria: February 24, 2025
• Last Verified: February 1, 2025

More Information

Terms related to this study

• Keywords: Machine Learning; POC Ultrasound; POC D-dimer
• Additional Relevant MeSH Terms: Vascular Diseases; Cardiovascular Diseases; Embolism and Thrombosis; Thrombosis; Venous Thrombosis; Molecular Mechanisms of Pharmacological Action; Antifibrinolytic Agents; Fibrin Modulating Agents; Hemostatics; Coagulants; Fibrin fragment D
• Other Study ID Numbers: 682139

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED
• IPD Plan Description: The plan to share IPD with other researchers are not yet decided because it has not yet been discussed.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No