Electrophysiological Surrogates and Statistical Rigor in Dose-Response Trials: Comment on “Early Detection and Temporal Progression of Hemidiaphragmatic Paralysis After Supraclavicular Brachial Plexus Block with 20 mL vs 30 mL Ropivacaine” [Letter]

Dear editor

We read with interest the article by Bao et al1 comparing the effects of 20 mL versus 30 mL of 0.375% ropivacaine on hemidiaphragmatic paralysis (HDP) following ultrasound-guided supraclavicular brachial plexus block (SCBPB). The authors are commended for their serial electrophysiological assessments and for addressing a clinically relevant dose-response question. However, several statistical and methodological issues warrant careful consideration, as they may affect the validity and generalizability of the conclusions.

First, the sample size calculation appears inadequately justified and may have led to underpowered subgroup or time-point comparisons. The authors based their power analysis on an assumed HDP incidence of 41% versus 68% from their previous study, but did not provide the effect size (eg, Cohen’s h) or the within-subject correlation structure for repeated measures. More importantly, the final analysis included only 70 patients (35 per group) instead of the planned 82, with 8 excluded post-randomization. Although exclusions were based on predefined technical criteria, the authors did not perform an intention-to-treat (ITT) analysis or sensitivity analyses to assess whether these exclusions influenced the primary outcome. In randomized trials, per-protocol analyses are prone to bias, particularly when exclusions are unevenly distributed or related to block success, as occurred here (2 vs 1 failures in Groups A and B, respectively).2 We suggest that the authors re-analyze the data using ITT principles and report the results alongside their current findings.

Second, the use of generalized estimating equations (GEE) for the primary outcome is appropriate, but several key model details are omitted. The authors did not specify the working correlation structure (eg, exchangeable, AR-1) used to account for repeated measurements within individuals, nor did they report the quasi-likelihood under independence model criterion (QIC) for model selection. Additionally, while time was treated as a categorical variable, the authors did not test for linear or nonlinear trends in HDP incidence over the 30-minute window. Given that the incidence plateaued by 20 minutes, a piecewise or spline model might have provided a more parsimonious and informative description.3 Without these details, readers cannot assess the robustness of the time-effect estimates.

Third, the primary endpoint-a ≥50% reduction in diaphragmatic compound muscle action potential amplitude (Dia CMAP Amp)-is an electrophysiological surrogate, not a direct measure of respiratory compromise. The authors appropriately acknowledge this limitation, but they then overinterpret the clinical implications by recommending a specific monitoring window (5–20 minutes post-block) without linking the electrophysiological threshold to clinically meaningful outcomes such as dyspnea, oxygen desaturation, or changes in forced vital capacity. In fact, only 14.3% of patients with HDP reported dyspnea, and only one patient developed desaturation. The low concordance between the surrogate endpoint and clinical symptoms raises the question of whether a ≥50% CMAP reduction is clinically meaningful in this healthy population. Boney et al, in developing the COMPAC core outcome set for perioperative trials, emphasized that surrogate endpoints require careful validation and should be interpreted alongside patient-centered outcomes to ensure clinical applicability.4 We suggest that the authors present a receiver operating characteristic (ROC) analysis to evaluate the optimal CMAP reduction threshold for predicting respiratory symptoms, or at least report the sensitivity and specificity of the 50% cutoff.

Fourth, the handling of missing data and protocol deviations is not fully transparent. Eight patients were excluded after randomization, including two with block failure in Group A and one in Group B. Block failure was defined as the need for supplemental blocks, opioids, or conversion to general anesthesia-events that could themselves be influenced by the volume of local anesthetic. Excluding these patients may have selectively removed those with inadequate blockade, thereby biasing the HDP incidence estimates. We recommend that the authors perform a sensitivity analysis in which block failure is treated as a treatment failure (ie, imputed as HDP or as missing not at random) to evaluate the stability of their conclusions. Recent guidance on estimands in anesthesia trials stresses that intercurrent events such as rescue analgesia should be prespecified and incorporated into the primary analysis framework.5

Finally, the authors report that the group-by-time interaction was not statistically significant (P = 0.072 for CMAP amplitude), yet they interpret the parallel declines as evidence that higher volume deepens block without extending its spread. This conclusion is speculative without a formal test of non-inferiority or equivalence in the rate of decline. Moreover, the absence of a significant interaction does not prove homogeneity of treatment effects across time; it only indicates insufficient evidence to reject the null hypothesis, which may be due to limited sample size. We suggest that the authors report the confidence intervals for the interaction term and discuss the clinical equivalence margins that would be relevant for interpreting this finding.

In summary, while this study provides valuable exploratory data on the temporal course of phrenic nerve block after SCBPB, the current analysis would benefit from greater methodological rigor, including ITT analysis, clearer model specification, sensitivity analyses for missing data, and a more cautious interpretation of surrogate endpoints. We appreciate the authors’ contribution and hope these comments help strengthen the evidence base for perioperative respiratory monitoring.

Disclosure

The authors report no conflicts of interest in this communication.