Modeling and Application of Triple Drug Response Surface Models

Triple Drug Response Surface Modeling for Patients Receiving Airway Managements

Demand for anesthesiologists outside the operating rooms is increasing. Surgeons, radiologists, endoscopists and other interventionists are performing procedures with greater complexity, sometimes accompanied by greater pain and therefore require some sedation. This growing need place a pivotal role on careful handling the delivered drugs. Specifically, the investigators wish to know how different classes of drugs interact in order for us to titrate the effects more precisely. Early studies used isobolograms and concentration effect curves as their tools but these methods are limited and incapable of making continuous bedside monitoring. Researchers integrated these methodologies mathematically and developed response surface models. It's becoming a very convenient tool to assess drug interactions. Drug interactions are visualized with a three dimensional graph, or a surface. Users will only need the calculated drug concentrations and a predicted loss of response probability will be given after a model is built. The process of model construction is complex and time demanding process. Our team has successfully built a dual-drug model for midazolam and alfentanil and several works have been published in renowned anesthesia journals. Dual-drug models are simple but their use is very limited. The investigators often use more than two drugs during practice and a three-drug model will be a great leap to monitor clinical pharmacodynamics. The investigators will collect vital signs, anesthestic depth (BIS=bispectral index, analgesia nociception index (ANI)), drug dosing (propofol, midazolam and alfentanil), and the MOAA/S (modified observer's assessment/alertness score) scores in patient who are receive routine general anesthesia under laryngeal mask or sedation for TEE (transesophageal echocardiography) examinations. Patient's consent will be obtained prior to enrollment. These recordings will be pooled into computer program for model construction. A novel model in the field of anesthesia was chosen and modified. As pioneers in this field in Taiwan, the investigators plan to perform a series of analysis using a novel model the investigators have built to look into detail on how drugs interact with each other. A safety boundary to avoid excessive respiratory depression can be drawn by the model. The main goal is to provide sedation that gives precision, patient comfort, rapid return of consciousness and safety based on the triple-drug response surface model.

This is an observational study of the routine clinical practice with no specific additional interventions required.

Data acquisition (TEE patients)

1. After screening for eligible patients, protocol and study details will be thoroughly explained to them.
2. Strict fasting protocols were followed. A 22-gauge intravenous catheter was secured for drug administration. Each patient received standard anesthetic care monitors comprised of ECG (electrocardiography), oxygenation saturation (SpO2) and NIBP (non-invasive blood pressure). ECG and SpO2 (oxygenation saturation) were monitored continuously and NIBP were measured every 5 minutes. Supplemental oxygen (3~5 L/m) was given via nasal cannula, and SpO2 (oxygenation saturation) maintained above 90%. Intravenous bolus midazolam, alfentanil and propofol were given according to the anesthesiologist's preferences. Reference dosage range for midazolam is 0~5mg, alfentanil 0~1000mcg and propofol 0~60mg. Anesthetic depth was monitored with BIS and analgesia nociception index (ANI). BIS and ANI (analgesia nociception index) were for monitoring purposes only.
3. Instrumentation began after loss of response as evaluated by the anesthesiologist with the modified observer's assessment/alertness scale (MOAA/S). Intolerable desaturation was managed with mask ventilation or insertion of a nasal airway. An MOAA/S score lesser than 2 was considered loss of response (LOR) and fit for instrumentation. MOAA/S score would be recorded at induction, instrumentation, and emergence phase. Recordings will be added if patient had a MOAA/S score greater than 2. ANI (analgesia nociception index) and forehead BIS monitors will be used as an adjunct to assess anesthetic depth. Additional drug boluses were given if the patient expresses pain or excessive movements observed.
4. At the end of the procedure, the patient was observed until return of consciousness (MOAA/S > 5). The patients will be randomly divided into a model training group and a validation group after data acquisition by computer randomization using computer clock as seed.

Data acquisition (Laryngeal mask patients)

1. After screening for eligible patients, protocol and study details will be thoroughly explained to them.
2. Strict fasting protocols were followed. A 22 or 20-gauge intravenous catheter was secured for drug administration. Each patient received standard anesthetic care monitors comprised of electrocardiography (ECG), SpO2 (oxygen saturation) and non-invasive blood pressure (NIBP). ECG and SpO2 (oxygen saturation) were monitored continuously and NIBP were measured every 5 minutes. Intravenous fentanyl and propofol were given according to the anesthesiologist's preferences. Reference dosage range for fentanyl 0~150mcg and propofol 0~200mg. Anesthetic depth was monitored with BIS and ANI (analgesia nociception index). BIS and ANI were for monitoring purposes only.
3. Instrumentation began after loss of response as evaluated by the anesthesiologist with the MOAA/S score. A MOAA/S score lesser than 2 was considered LOR (loss of response) and fit for instrumentation. After LMA placement, sevoflurane was started at 2~3% with a fresh gas flow at 1~6 L/m. The end-tidal sevoflurane concentration would be maintained above 0.7 minimal alveolar concentration to avoid awareness. MOAA/S score would be recorded at induction, instrumentation, during skin incision/closure and emergence phase. Recordings would be added if patient had a MOAA/S score greater than 2. ANI (analgesia nociception index) and forehead BIS monitors would be used as an adjunct to assess anesthetic depth. Additional drug boluses were given if the patient expresses pain or excessive movements observed.
4. At the end of the procedure, LMA was removed if the patient was breathing smoothly. The patient was observed until return of consciousness (MOAA/S > 5). The patients will be randomly divided into a model training group and a validation group after data acquisition by computer randomization using computer clock as seed.

Model building, assessment and validation (both groups)

1. Collected data were fed to a pharmacokinetic simulation software (TIVA trainer Version 9.1) to calculate second-by-second plasma and effect-site drug concentration changes for all three drugs. The training patient group is used for model training. The bootstrap technique is used with 2000 iterations. Model fit was optimized using -2 Log likelihood (-2LL).
2. MOAA/S, BIS and ANI (analgesia nociception index) models would be constructed.
3. The results are validated with the validation patient data to confirm its clinical utility. Receiver operating characteristics (ROC) curve analysis is used to assess the quality of model prediction. ROC and area under the curve (AUC) will be compared between the training and validation group.

Inclusion Criteria:

• Patients aged between 20 and 85
• Scheduled for transesophageal echocardiography (TEE) examination or surgery under laryngeal mask airway (LMA) depending on study arm.
• American Society of Anesthesiologists (ASA) physical status I to III.

Exclusion Criteria:

TEE arm:

• hearing impairment
• neurologic or behavioral disorders
• habitual sedative use
• alcoholism
• allergy to midazolam, alfentanil or propofol
• resting room air SpO2 < 90%.
• History of upper airway tumors

LMA arm

• hearing impairment
• neurologic or behavioral disorders
• habitual sedative use
• alcoholism
• allergy to sevoflurane, fentanyl or propofol, resting room air
• SpO2 < 90%
• poor dentition
• non-fitting LMA
• LMA insertion attempt greater than 2.