Aligning prediction targets before ranking diaphragm ultrasound parameters in AECOPD

To the Editor,

We read with interest the systematic review and meta-analysis by Zhou et al. evaluating diaphragmatic ultrasound for predicting weaning outcomes in mechanically ventilated patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) [1]. This disease-specific synthesis is clinically valuable, and the authors appropriately caution against overinterpreting apparent prognostic performance derived from heterogeneous studies and post hoc thresholds. Two additional considerations may further clarify how these findings should inform bedside decisions during ventilator liberation.

First, variation in the definition of weaning failure may affect more than statistical heterogeneity; it may change the prediction target itself. The included studies defined failure using outcomes ranging from spontaneous breathing trial (SBT) failure to reintubation, non-invasive respiratory support, tracheostomy, or death, while diaphragm ultrasound was performed at different points before, during, or after the SBT [1]. Although the authors explored whether inclusion of SBT failure explained heterogeneity by meta-regression, these outcomes remain clinically distinct even if no study-level statistical association is detected.

Weaning and extubation are sequential but separate processes [2]. SBT failure primarily reflects inability to tolerate unsupported or minimally supported breathing, whereas extubation failure after a successful SBT can additionally result from airway obstruction, ineffective cough or secretion clearance, cardiac dysfunction, neurological impairment, and other mechanisms [2, 3]. Consequently, when diaphragm ultrasound is performed during an SBT and SBT failure is included in the reference outcome, the measured ultrasound parameters may partly reflect the patient’s physiological response to the ongoing trial rather than purely predict a subsequent weaning outcome. This differs from using diaphragm ultrasound in a patient who has already passed an SBT to predict whether liberation will remain successful after extubation. The former more closely reflects SBT tolerance and readiness for liberation, whereas the latter addresses the durability of liberation after extubation.

Pooling these states under a single “weaning failure” outcome may therefore combine different clinical questions rather than merely introduce heterogeneity. Where sufficient data exist, a decision-point analysis separating prediction of SBT failure from prediction of post-SBT extubation failure would be informative. If quantitative stratification is not feasible, presenting these evidence streams separately would still clarify where in the liberation pathway diaphragm ultrasound provides prognostic information.

Second, the finding that diaphragmatic excursion (DE) yielded the highest pooled area under the curve should remain a descriptive comparison rather than evidence of parameter superiority. DE, diaphragm thickening fraction (DTF), diaphragmatic rapid shallow breathing index (D-RSBI), and diaphragmatic contraction velocity were synthesized from 16, 13, 8, and 4 studies, respectively [1]. These parameter-specific estimates therefore arise from partly different evidence sets with potentially different case mix, SBT protocols, outcome definitions, measurement timing, and thresholds. Comparing separately pooled AUCs cannot establish that DE performs better than DTF or D-RSBI in otherwise comparable patients.

This distinction is particularly relevant because a previous comparative meta-analysis published in Critical Care formally compared respiratory muscle assessments using hierarchical models and direct comparisons restricted to studies evaluating competing tests within the same patients [4]. In those direct comparative analyses, DTF showed significantly greater predictive accuracy than DE [4]. This finding in a broader ICU population does not contradict the AECOPD-specific estimates reported by Zhou et al.; rather, it illustrates why parameter ranking requires comparative evidence rather than parallel synthesis of partly different study sets.

An informative extension would therefore be a paired analysis of AECOPD studies that measured DE and DTF, or other competing indices, in the same participants under the same reference standard. If too few such studies are available, the conclusion could more precisely state that DE currently has the highest pooled estimate among separately synthesized evidence bases, without implying that it is the optimal diaphragmatic ultrasound parameter.

These considerations refine rather than diminish the contribution of Zhou et al. The clinically relevant question may not simply be which diaphragm ultrasound parameter has the highest pooled accuracy, but which parameter adds useful information at which decision point in ventilator liberation. Aligning the prediction target first and comparing parameters under equivalent clinical conditions second would make this growing evidence base more directly actionable.

Author contributions

Li-Xia Xu conceived the commentary and drafted the manuscript. All authors critically revised the manuscript and approved the final version.

Funding

No funding was received for this work.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.