Outcome-Driven Treatment Delivery & Personalized Medicine
Applying an epidural to a patient.
CHOT researchers focused on an evidence-based approach to treatment delivery. Two teams of investigators were involved: 1) a personalized treatment design team for diabetic patients, and; 2) an optimizing epidural analgesia procedures team.
The potential consequences of failed or misplaced epidural needles are well known to obstetric anesthesiologists. Inadvertent intravenous injection of local anesthetic into a vein in the epidural space can lead to seizures or fatal cardiac arrhythmias. Equally worrisome are inadequate epidural blocks leading to complications during caesarian sections.
This study sought to establish and quantify the safety and efficacy of a large-dose needle-based epidural technique in obstetric anesthesia. Through systems modeling and predictive analytics, a safe and quickly effective epidural dose is established that can then be administered through the epidural needle prior to the insertion of the epidural catheter.
The results indicate that a needle-based approach is 22% faster and more dose-effective (18% less drug) in achieving comparable sensory levels than the traditional catheter-based approach. The findings also suggest that injecting large doses (up to 20mL) in the epidural space through the epidural needle is generally safe and results in good outcome for the patients.
The end product of this work is a decision support system that can identify patients-at-risk of complications. It also highlights the best practice clinical practice guideline (CPG) that allows for standardized, safer, and more cost-effective epidural delivery with minimum complications.
The cumulative costs of approximately four million annual births are well over $50 billion. This research demonstrates that it may be possible the use fewer drugs to achieve desired sensory responses with minimal hypotension. Reducing hypotension may also lead to safer process and possibly better long term outcomes. Providing quality of care with minimal complication is of paramount importance. Effective training of our new physicians (anesthesiologists) means more treatment success and improved efficiency. Given that physician time is expensive, this work thus reduces wastes in physician time and maximizes their productivity.
The management of gestational diabetes mellitus (GDM) requires close monitoring of patients’ blood glucose levels while clinicians experiment with dosing based on a combination of clinical guidelines and their experience and judgment. However, conflicting guidelines and wide variation in practices can result in less that optimal care.
A challenge in diabetes management comes from the fact that different patients have different doseresponses and different disease progression characteristics. Hence, a personalized treatment plan tailored specifically to the patient’s unique dose-effect characteristics may be more effective and efficient than current trial-and-error approaches.
In this project, CHOT researchers designed a novel outcome-based decision support tool that couples a predictive treatment-effect model with a planning optimization model. Specifically, a predictive model first uncovers treatment effects based on pharmacokinetic and pharmacodynamics (PK/PD) analysis of anti-diabetic drug dosages. Blood glucose levels are then recorded (self-monitored blood glucose) during the titration period of each patient. This information is then incorporated within a personalized planning model for optimal treatment. The decision support tool makes possible continuous learning for each patient as new treatment outcomes are recorded.
Tested on a group of 200 patients, using the first 2-3 weeks of treatment to establish the predictive drug effect, results indicate that the optimized treatment plan may offer improved glycemic control with lower drug usage compared with current practice.
The predictive PK/PD treatment-effect model becomes more precise as outcome data accumulate. Most PK/PD models require drug concentration levels in the blood but these are not generally measured. This new approach seems to overcome this obstacle while establishing more direct relationships between drug dosages and drug effects. Incorporating this information within a treatment planning optimization model allows clinicians to tailor outcomes and medication regimens to the individual patients’ specific needs. Over time, this approach may lead to better treatment decisions and possibly improved outcomes.
In summary, the PK/PD treatment-effect model is a mechanism-based that captures each patient’s underlying glucose dynamics and drug effects. By doing so it offers predictive estimates of dose glucose response characteristics. It captures disease progression over wider treatment horizons. This helps assure that complying patients have the adequate glycemic control that is necessary for safe drug delivery. Because it uses only drug dosage and self-monitored blood glucose levels (SMBG) that are hopefully recorded accurately by patients at home, for compliant patents it is readily implementable with current clinical and patient practice. Last and most importantly, the model is tailored to each individual patient to obtain a personalized dose response and disease progression. This predictive information is then incorporated into a mathematical programming-based treatment model that optimizes their glycemic control and drug dosage.
The end product is a real-time clinical decision support system that enables clinicians to tailor treatment design to the needs of the patients. It should help clinicians make better treatment decisions.
For more information, contact Eva Lee at the Georgia Institute of Technology, firstname.lastname@example.org, Bio https://www.isye.gatech.edu/users/eva-lee, 404.894.4962.CHOT-2016.pdf