Neuromuscular Blocking Agents During Electroconvulsive Therapy: Was the Crossover Design Appropriate and Valid?
This is a “composite” study that can be divided into 3 distinct studies. The first and second are dose-finding studies to identify the minimum effective doses of succinylcholine and rocuronium, respectively. On the other hand, the third is a comparative study of the duration of the neuromuscular blockade.
Unlike the comparative study, the dose-finding studies did not benefit from the crossover design. In each dose-finding study, the optimal dose for each patient was estimated, and these individual estimates from multiple patients were then used to calculate the median to generate the 50% effective dose for each neuromuscular blocking agent (NMBA). The value of the crossover design only appeared when the 2 NMBAs were compared with each other, which did not occur in the dose-finding studies. It would be more appropriate if the dose-finding studies were performed on a cohort of patients, and then the resulting minimum effective dose of each NMBA was utilized in a comparative study on a different cohort.
Ideally, paired analyses should be performed for crossover studies. It is not clear whether unpaired or paired analyses were performed. During the description of sample size calculations, “Wilcoxon test” was mentioned. Mostly the authors were referring to Wilcoxon signed-rank test, a paired test. However, this causes confusion with Wilcoxon rank-sum test, an unpaired test. Eventually, comparison of the measured variables was done using the Welch t test, an unpaired test. Even, the usage of the Welch t test seems to be strange because the Welch t test is used when the 2 samples have unequal variances and unequal sample sizes, which is not expected because of the crossover design and the sample size of >30.
One of the essential prerequisites of crossover studies is the absence of change in the condition being examined during the course of the study.2 This cannot be guaranteed in the current study because it is known that changes in seizure threshold occur during the course of ECT. It was not mentioned whether or not the seizure threshold was measured or calculated at the beginning of the study or during the study. In addition, no ictal electroencephalogram indices were utilized as an indicator of changes in the relative stimulus intensity.3
Regarding the ECT parameters mentioned in Table 1 of the current study, pulse width had the unit of mA (milliamperes), which is not correct because the pulse width is typically reported in milliseconds. In addition, mA is a unit for current amplitude, not time. Also, the range for pulse width (mA) was 0 to 2. A minimal value cannot be zero for either pulse width or current amplitude. The dose of ECT was expressed as the total energy of the stimulus train (in joules). This was common before 1980s; however, it was replaced by total charge (in millicoulombs) for reasons that are beyond the scope of this letter.