Study design
This randomized assessor-blind controlled trial with two parallel arms was conducted at Peking University First Hospital. Prior to enrollment, the study protocol was approved by the local Ethics Committee (2019-222) and was registered prospectively at chictr.org.cn (ChiCTR-IPR-1900027585, Principal investigator: Hong Zhang, Date of registration: November 19, 2019). Written informed consent was obtained from all participating trainees.
Participants
Anesthesia residents in a clinical rotation in anesthesiology for more than 3 years were screened for eligibility. Residents who had previous systematic experience with ultrasound-guided nerve block or who were unwilling to participate in this study were excluded.
Prior to training, we sent a questionnaire survey to collect baseline data, including sex, age, handedness, educational background, years on rotation in anesthesiology, and previous experience with ultrasound-guided procedures.
Randomization, training course and assessment
Trainees were randomly assigned within blocks of rotation years into the workshop group (WG) or the control group (CG) according to a computerized random number generator by an independent biostatistician (Xue-Ying Li) in a 1:1 ratio using the SAS 9.3 statistical package (SAS Institute, Cary, NC, USA).
All trainees received didactic lectures on the basics of ultrasound (30 minutes), transverse abdominis plane block (TAP) (30 minutes), thoracic paravertebral block (PVB) (30 minutes), brachial plexus block (BPB) (30 minutes), and femoral nerve block (FNB) (30 minutes). To minimize fatigue, there were 10-minute breaks between the lectures. The acquisition and recognition of sonographic images of the nerves and plane layers were emphasized. Additionally, the trainees in the WG received five fifteen-minute one-to-one interactive simulation-based training sessions on the following topics: TAP, PVB, BPB, FNB and UGRA skills. The sessions were facilitated by five senior anesthesiologists (YTL, XL, HK, DH and FZ) who designed and prepared the sessions collectively. The workshop on UGRA skills used a silicon model for interactive practice of needle visualization for single-injection and continuous UGRA, while the other workshops used human volunteers for ultrasound scanning and sonographic structure recognition. The four human volunteers who remained throughout the training were recruited and consented to participate in each session to ensure the same conditions for each trainee. Real-time feedback with skill adjustments on UGRA was immediately provided during the workshop period.
The knowledge and skills of the trainees in both groups were subsequently assessed by five assessors (HZ, FC, ZZX, ZTM, ZML) who had not been involved in the previous training and who were completely blinded to the groupings. They were in charge of five assessment stations (TAP, PVB, BPB and FNB ultrasound scanning in live human models and UGRA procedure in porcine meat models), and prior to any assessments, consensus training was conducted to instruct these assessors to execute an identical evaluation standard for all trainees. One of the authors (HZ) designed 100 multiple choice questions (MCQ) focusing on the five sections mentioned in the didactic lectures to create a 60-minute written examination.
The porcine meat models were established as follows: Model A, porcine meat containing distinguishing fascial layers whose corresponding ultrasound image was similar to the layers in the TAP and PVB (Figure 1); Model B, porcine meat with embedded bovine tendon whose corresponding ultrasound image was similar to the nerve in BPB and FNB (Figure 2), as described by Xu and colleagues [18]. The trainees were instructed to successfully perform a short axis, in-plane ultrasound-guided target injection using the porcine meat models. Identical 22G facet beveled, nonechogenic nerve block needles (Stimuplex D; B Braun, Germany) were used for both the training and the examination. An ultrasound machine with a high frequency 5-13 MHz linear transducer was used (Logiq-e, GE Inc., USA). The ultrasound images, the hand maneuvers of the trainees and the transducer movements were continuously recorded by a learning management solution system (SimCapture, B-line medical, USA) through data records and a video camera (Figure 4). Two evaluators (FZ and ZML) assessed the trainees using video analysis.
The assessment outcomes consisted of three parts: (1) the MCQ in the written examination, which focused on basic ultrasound theory, sonographic anatomy structure recognition and knowledge of clinical applications for four routine nerve blocks (TAP, PVB, BPB and FNB). Each topic included 20 questions. (2) Drawing the anatomy and ultrasound image acquisition and interpretation on four routine nerve blocks (TAP, PVB, BPB and FNB) in live human models. The drawing score for four nerve blocks ranging from a maximum score of 7 to 17 was calculated by summing the anatomical parameters listed in Appendix 1 section 1. The sonographic proficiency score for the four nerve blocks listed in Appendix 1 section 2 was calculated to evaluate proficiency, with a maximum score ranging from 6 to 18. The nerve recognition score for BPB and FNB listed in Appendix 1 section 3 was calculated to identify the terminal branches of the related nerve trunk or plexus. (3) The time taken for ultrasound-guided target injection, the cumulative score of errors and the worst sonographic image score in porcine meat models.
The time taken to correctly complete the nerve block, defined as the time from picking up the transducer to the successful deposition of injectate on two target sites, including 1 ml of saline in a designed layer Model A and 1 ml of saline at the 12 o’clock and 6 o’clock positions relative to the nerve in a designed nerve Model B. The cumulative score of errors defined as the errors detected during the procedure is listed in Appendix 2. The image quality score was defined as the worst image of the simulated nerve in Model A or Model B during the procedure, scored numerically (1=unsatisfactory, 2=poor, 3=satisfactory, 4=outstanding) in Appendix 2. The means of the above results by the two evaluators were calculated and recorded.
The primary outcome was the time taken to correctly complete the nerve block in the porcine meat model. The secondary outcomes included the following: (1) the MCQ score; (2) the score of drawing the anatomy; (3) the sonographic proficiency score; (4) the nerve recognition score; (5) the total score of every participant in the four live human model stations; (6) the cumulative score of errors; and (7) the worst sonographic image quality of the target nerve achieved.
Statistical analysis
Sample size estimation
Previous studies reported the mean length of time to perform the tasks at 37.63 seconds for simulation training vs. 93.83 seconds for nonsimulation training [19]. With the significance and power set at 0.05 (two-sided) and 90%, respectively, the sample size required to detect differences was 14. Considering a drop-out rate of approximately 25%, we planned to enroll 20 trainees. The sample size calculation was performed with PASS 11.0 software (Stata Corp. LP, College Station, TX).
Data analysis
Normally distributed continuous variables are expressed as the mean ± SD and were compared using a two-tailed Student’s t-test. Nonnormally distributed continuous variables and ordinal data are expressed as the median (interquartile range) and were analyzed using the Mann-Whitney U test. Categorical variables were expressed as case numbers and were analyzed by Chi-square or Fisher’s exact test. A two-sided P value of less than 0.05 was regarded as statistically significant. All statistical analyses were performed with SPSS statistical package version 25.0 (IBM Corp. Armonk, NY, USA).