Ultrasound-detected diaphragmatic dysfunction after adult cardiac surgery: an ascertainment-dependent phenotype—a systematic review and meta-analysis

Introduction

Background

Diaphragmatic dysfunction (DD) is a well-recognised complication of adult cardiac surgery, encompassing a spectrum from transient, subclinical contractile impairment to overt hemidiaphragmatic paralysis with respiratory failure (1,2). Bedside ultrasound has become the reference modality for diaphragm assessment because it is non-invasive, repeatable and feasible in ventilated and spontaneously breathing patients, quantifying both motion (excursion) and contractile activity (thickening fraction) (1,2). Plausible mechanisms after cardiac surgery are several and partly laterality-specific, including phrenic nerve cold injury from topical cardiac cooling, devascularisation or stretch of the phrenic nerve during internal mammary artery harvest, and the systemic inflammatory response to cardiopulmonary bypass (1) (Figure 1).

Figure 1 Proposed laterality-specific mechanisms of phrenic/diaphragmatic injury during cardiac surgery. ROS, related infiammation & oxidative stress; CPB, cardiopulmonary bypass.

Because cardiac surgery patients require mechanical ventilation, diaphragmatic function may be further affected by ventilation-related processes: ventilator-induced lung injury (VILI) (3), patient self-inflicted lung injury (P-SILI) during vigorous spontaneous effort (4), and ventilator-induced DD (VIDD), the contractile impairment and atrophy that follow controlled ventilation and diaphragm inactivity (5). These mechanisms are conceptually distinct from perioperative phrenic-nerve and mechanical injury but may compound it, and ultrasound cannot by itself distinguish their respective contributions.

Rationale and knowledge gap

Despite a consistent qualitative message, the quantitative literature is strikingly discordant: reported prevalence of post-cardiac-surgery DD spans from roughly 3% to more than 70%, and the association of DD with hard respiratory outcomes is inconsistent across reports, being negligible in some screen-detected cohorts yet strong in cohorts selected for persistent or symptomatic dysfunction (6-8). A recent meta-analysis confirmed that ultrasound-detected DD is associated with postoperative pulmonary complications, but it pooled cardiac, thoracic and abdominal surgery together and quantitatively combined only six studies, leaving the cardiac-surgery-specific picture, and the source of the prevalence variability, unresolved (9). We hypothesised that the apparent disagreement is not primarily biological but methodological, reflecting differences in ascertainment (systematic screening of every patient versus imaging only clinically symptomatic patients), timing of assessment (early transient versus persistent dysfunction) and diagnostic threshold (excursion versus thickening-fraction cut-offs). Standardising how DD is detected, timed and defined—and thereby identifying the persistent, symptomatic phenotype that plausibly carries clinical weight—is the contribution we pursue here; we use the language of a treatable trait to argue for systematic detection and phenotype-based monitoring, while acknowledging that no intervention trials yet exist to show that such stratification changes outcomes (10,11).

Objective

We therefore performed a cardiac-surgery-specific systematic review and meta-analysis to (I) quantify the pooled prevalence of ultrasound-detected DD and formally test whether ascertainment explains its heterogeneity, and (II) estimate the association between DD and reintubation, postoperative pneumonia and non-invasive ventilation (NIV). We present this article in accordance with the PRISMA reporting checklist (available at https://shc.amegroups.com/article/view/10.21037/shc-2026-0021/rc) (12).

Methods

Registration

The review was registered in PROSPERO (CRD420261430128). No important amendments were made.

Eligibility criteria

We included studies that (I) enrolled adults (≥18 years) undergoing cardiac surgery; (II) assessed the diaphragm by ultrasound (excursion, thickening/thickening fraction, or M-mode sniff test); and (III) reported the prevalence of DD and/or its association with at least one respiratory outcome (reintubation, postoperative pneumonia, NIV, tracheostomy, duration of mechanical ventilation, mortality). Both prospective and retrospective observational designs and randomised trials were eligible. We excluded mixed-surgery cohorts in which cardiac-surgery patients could not be separated, thoracic or abdominal surgery, paediatric populations, studies without ultrasound assessment of the diaphragm, and reviews, editorials, letters and protocols. Studies were grouped a priori into two syntheses: a prevalence synthesis (all studies reporting DD prevalence) and association syntheses (studies reporting DD status against a respiratory outcome); a study could contribute to one or both.

Information sources and search

We searched PubMed (title/abstract) up to June 2026; no other database was searched, and no language restriction was applied. The search combined diaphragm terms (“diaphragm”, “diaphragmatic”, “phrenic”), ultrasound terms (“ultrasound”, “ultrasonography”, “sonographic”, “echography”) and cardiac-surgery terms (“cardiac surgery”, “cardiac surgical”, “cardiopulmonary bypass”, “CABG”, “valve surgery”, “sternotomy”) with the Boolean operators AND/OR; the full string is reported in the Appendix 1 and in the Figure 2 footer. Reference lists of included studies and relevant reviews were screened for additional records. Reliance on a single database is acknowledged as a limitation.

Figure 2 PRISMA 2020 flow diagram (PubMed search; 90 records → 7 included). The full search string is reproduced in the footer; an editable Word/PowerPoint version is provided. NIV, non-invasive ventilation.

Selection and data extraction

Two reviewers independently screened records by title/abstract and then at full text against the eligibility criteria, with disagreements resolved by consensus with a third reviewer; the same two reviewers independently extracted data using a standardised form. For each study we extracted the design, country and centre, surgical population, ultrasound technique and DD definition/threshold, timing of assessment, DD prevalence, laterality, and the 2×2 counts or summary statistics for each respiratory outcome. One non-English report by Zhu et al. (13) was identified but excluded because only an English abstract was available; all included studies are full peer-reviewed reports. Methodological quality was appraised independently by two reviewers using the Newcastle-Ottawa Scale (NOS) for observational studies.

Ascertainment classification

Each study was classified a priori, on the basis of its stated design, as using systematic screening (protocolised diaphragm ultrasound in all consecutive patients) or symptomatic/clinically triggered ascertainment (ultrasound performed only in patients with clinical signs prompting investigation, e.g., chest-radiograph elevation or failure to wean). This classification was the pre-specified primary moderator for the prevalence analysis.

Statistical analysis

Analyses were performed in R 4.3 with the metafor package (version 4.4.0). Prevalence was pooled as proportions after logit transformation under a random-effects model fitted by restricted maximum likelihood (REML), back-transformed with 95% confidence intervals (CIs). For dichotomous outcomes we computed study-level odds ratios (ORs) and pooled them with a random-effects (REML) inverse-variance model; a 0.5 continuity correction was applied to zero cells (6). Heterogeneity was quantified with the I² statistic and explored through the pre-specified ascertainment subgroup analysis; because few studies were available per covariate, formal meta-regression was not undertaken. Robustness was examined with leave-one-out and Baujat analyses for the reintubation and pneumonia outcomes, and small-study effects with a contour-enhanced funnel plot for prevalence (Figure S1); formal tests of funnel asymmetry were not performed where fewer than 10 studies were available, as they are underpowered. Two-sided P values are reported.

Results

Study selection and characteristics

The PubMed search identified 90 records. After title/abstract screening (70 excluded) and full-text assessment of 20 reports (13 excluded with reasons), seven studies met the inclusion criteria (Figure 2). All seven contributed to the prevalence synthesis; five, three and three studies respectively contributed to the reintubation, pneumonia and NIV association analyses. The included studies (6-8,14-17) comprised prospective and retrospective cohorts performed between 2020 and 2026 across European, North American and Asian centres, ranging in size from 67 to 3,577 patients, and used heterogeneous DD definitions (excursion <1 or <2 cm; thickening fraction <20%; M-mode sniff test below sex-specific limits). Study-level characteristics, DD definitions and outcome data are summarised in Tables 1-3, and risk-of-bias judgements in Table 4.

Prevalence of DD and the role of ascertainment

The overall pooled prevalence of ultrasound-detected DD across the seven studies was 22% (95% CI: 8–47%); given extreme heterogeneity (I²=99.0%) this single estimate is not clinically interpretable and is reported only to motivate the subgroup analysis (Figure 3). The pre-specified ascertainment moderator accounted for much of the between-study variation: pooled prevalence was 36% (95% CI: 17–61%) in studies that screened all patients systematically (6,8,14-16), compared with 5% (95% CI: 2–12%) in studies using symptomatic/clinically triggered imaging (7,17); substantial residual heterogeneity nonetheless remained within the systematic subgroup (Figure 3). The extremes of the literature were consistent with this framework: systematic post-rehabilitation screening reported DD in 131/185 patients (71%) (15), whereas a large retrospective study imaging only clinically suspected cases reported 272/3,577 (7.6%) (7). Early systematic studies further showed that much screen-detected DD is transient, peaking on the first or second postoperative day and largely recovering by day 3–7 (14), whereas persistent dysfunction beyond day 7 was uncommon (10/122, 8%) (8).

Figure 3 Forest plot of DD prevalence by ascertainment subgroup with study weights (systematic 36% vs. symptomatic 5%); proportions on the linear scale. CI, confidence interval; DD, diaphragmatic dysfunction; RE, random effects.

DD and respiratory outcomes

For reintubation (five studies; 6-8,16,17) the pooled OR was 3.07 (95% CI: 0.85–11.11; P=0.09; I²=73.2%), a non-significant trend that should be regarded as exploratory (Figure 4A). Influence analyses showed the estimate to be unstable: in a post-hoc leave-one-out analysis, omitting the persistent-DD cohort of Laghlam 2023—which enrolled only patients with DD persisting beyond postoperative day 7—yielded a significant pooled OR of 2.53 (95% CI: 1.41–4.54), and Baujat analysis identified the two symptomatic/persistent cohorts (7,8) as the most influential contributors (Figures S2,S3).

Figure 4 Association of DD with respiratory outcomes (odds ratios on a log scale, with study weights): (A) reintubation (exploratory); (B) postoperative pneumonia; (C) non-invasive ventilation (exploratory). CI, confidence interval; DD, diaphragmatic dysfunction; OR, odds ratio; REML, restricted maximum likelihood.

For postoperative pneumonia (three studies; 7,8,17) the pooled OR was 3.10 (95% CI: 2.32–4.15) with no detectable heterogeneity (I²=0.0%), the most consistent association in the review (Figure 4B). This association was robust in leave-one-out analysis, remaining significant when any single study was removed, including omission of Laghlam 2023 (OR 2.96, 95% CI: 2.20–3.98).

For NIV (three studies; 6,7,16) the pooled OR was 5.71 (95% CI: 1.70–19.21; P=0.005; I²=79.2%) (Figure 4C). This estimate is exploratory: the NIV indication differed across studies—rescue NIV (6), sequential post-weaning NIV (16) and any NIV (7)—and the large “any NIV” cohort of Laghlam 2021 dominates the pooled estimate [per-study OR 2.20 (0.55–8.75), 3.61 (1.07–12.21) and 14.81 (11.16–19.64), respectively].

Outcomes reported by only two studies were summarised narratively (Table 3): tracheostomy was more frequent with DD [9/272 vs. 10/3,305 (7); 1/72 vs. 1/105 (16)], and short-term mortality was not increased by DD, being numerically lower in the largest symptomatic cohort (2/272 vs. 97/3,305) (7) and nil among DD patients in the minimally invasive series (0/35 vs. 19/1,120) (17).

Surgical approach and laterality

Laterality was reported inconsistently. Where documented, left-predominant DD accompanied sternotomy and coronary surgery, consistent with left phrenic exposure during left internal mammary artery harvest and topical cooling (16); the source studies did not, however, provide patient-level correlation between individual mammary-artery harvest and the side of injury. In a single large single-centre series, minimally invasive mitral valve surgery through a right thoracotomy was independently associated with DD (3.0% overall; 1.8% after full sternotomy vs. 9.3% after the minimally invasive approach; adjusted OR 5.1, 95% CI: 2.37–11.0; P<0.001), with all minimally invasive cases right-sided (17). Because the right-sided signal derives from a single, non-externally-validated comparative study and may reflect centre-specific surgical practice, it is reported descriptively and treated as hypothesis-generating; further studies are warranted.

Discussion

Key findings

In this cardiac-surgery-specific synthesis of seven ultrasound studies, the more-than-twenty-fold spread in reported DD prevalence (approximately 3% to 71%) was largely, though not wholly, explained by ascertainment: systematic screening yielded a pooled prevalence of 36% and symptomatic imaging only 5%, with substantial residual heterogeneity remaining. Outcome risk did not track the headline prevalence figure but concentrated in the persistent, symptomatic phenotype. The association with postoperative pneumonia was the most robust and homogeneous (OR 3.10; I²=0.0%), whereas the reintubation and NIV associations were exploratory; the reintubation signal, in particular, was not significant overall and was sensitive to the persistent-DD cohort. These observations support standardising the ascertainment, timing and thresholds used to detect DD and—as a hypothesis for future testing—reframing it as an ascertainment-dependent treatable trait rather than a single dichotomous complication (Figure 5).

Figure 5 Diaphragmatic dysfunction after cardiac surgery as an ascertainment-dependent treatable trait (graphical synthesis).

Strengths and limitations

The principal strength is the deliberate restriction to cardiac surgery with an a priori ascertainment moderator, complemented by influence and sensitivity analyses that expose the fragility of the reintubation signal and by a formal risk-of-bias appraisal (Table 4). Several limitations should be acknowledged. First, the search was confined to a single database (PubMed); although supplemented by reference screening, this may have missed records indexed elsewhere and is an important limitation for a systematic review. Second, substantial residual heterogeneity remained even within the ascertainment subgroups, so the subgroup estimates should not be over-interpreted, and with few studies per covariate formal meta-regression of timing, threshold and population was not feasible. Third, the included studies were clinically and methodologically heterogeneous in DD definitions, thresholds and timing. Fourth, two cohorts originated from the same institution (Clinique Ambroise Paré) and may not be fully independent, although they are distinct cohorts contributing to different outcome nodes and were not double-counted within any pooled estimate (7,8). Fifth, the association analyses rest on few studies with sparse events and wide CIs and are therefore exploratory; certainty of evidence was not formally graded. Formal small-study-effect testing was not performed given the small number of studies.

Comparison with similar research

Our findings extend and refine the recent mixed-surgery meta-analysis by Jimenez-Santana et al., which pooled six studies across cardiac, thoracic and abdominal surgery and reported associations of DD with pulmonary complications (OR 2.99) and pneumonia (OR 5.41) (9). Restricting the analysis to cardiac surgery adds three specific things: a homogeneous, cardiac-specific pneumonia estimate (OR 3.10; I²=0.0%) with a more conservative point estimate; the explicit separation of ascertainment, timing and threshold effects that a mixed-surgery pool necessarily conflates; and the identification of the persistent/symptomatic phenotype as the locus of outcome risk. The narrative reviews on diaphragm ultrasound in cardiac surgery anticipated this clinical heterogeneity qualitatively (1,2); the present work quantifies it.

Explanations of findings

Two mechanisms reconcile the data. First, ascertainment and timing: systematic protocols capture a large reservoir of early, transient, mostly subclinical contractile impairment that recovers within days (14), whereas symptomatic imaging selects the smaller subset of persistent, clinically manifest dysfunction that drives reintubation and pneumonia (7,8). The leave-one-out finding for reintubation is consistent with this interpretation: the pooled association becomes significant when the small, influential persistent-DD cohort is removed, underscoring that clinically meaningful risk concentrates in persistent rather than transient dysfunction. Second, laterality: left-predominant DD after sternotomy or coronary surgery is consistent with left phrenic injury during internal mammary harvest and topical cooling (16), whereas the exclusively right-sided DD reported after right-thoracotomy minimally invasive mitral surgery (17) is a single-centre observation that requires external validation before any laterality-by-mechanism rule can be asserted.

Implications and actions needed

Clinically, DD should be actively sought in patients with difficult weaning, post-extubation respiratory distress or unexplained hypoxaemia, where a positive bedside study carries greater prognostic weight than an incidental screen-detected finding. For research and reporting, studies of post-cardiac-surgery DD should pre-specify and report their screening strategy, assessment timing and diagnostic threshold, distinguish transient from persistent dysfunction, and report laterality in relation to the surgical approach, so that prevalence figures become comparable and outcome estimates interpretable. We use “treatable trait” in this measurement-and-monitoring sense—a defined, quantifiable attribute that should be sought and tracked—and emphasise that no intervention trials currently exist; whether phenotype-based management improves outcomes is a hypothesis that future prospective studies should test (10,11).

Conclusions

Ultrasound-detected DD after adult cardiac surgery is common but profoundly ascertainment-dependent: its wide reported prevalence (approximately 3% to 71%) largely reflects how, when and with which threshold it is sought, although substantial residual heterogeneity remains, while its most consistent clinical correlate is an association with postoperative pneumonia concentrated in the persistent, symptomatic phenotype. Standardising ascertainment and reporting would make future estimates comparable. Prospective ultrasound cohorts already exist; what is now needed are prospectively registered studies with pre-specified, standardised screening timing and thresholds and explicit phenotyping to test whether identifying and monitoring this trait improves outcomes.

Acknowledgments

Use of AI: the authors disclose the use of a large language model-based AI assistant for literature-search support, data tabulation, generation of meta-analytic figures and language editing. All outputs and analyses were reviewed and verified by the authors, who take full responsibility for the content.

Footnote

Reporting Checklist: The authors have completed the PRISMA reporting checklist. Available at https://shc.amegroups.com/article/view/10.21037/shc-2026-0021/rc

Peer Review File: Available at https://shc.amegroups.com/article/view/10.21037/shc-2026-0021/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://shc.amegroups.com/article/view/10.21037/shc-2026-0021/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Fayssoil A, Mansencal N, Nguyen LS, et al. Diaphragm Ultrasound in Cardiac Surgery: State of the Art. Medicines (Basel) 2022;9:5. [Crossref] [PubMed]
2. Rizza V, Maranta F, Cianfanelli L, et al. Imaging of the Diaphragm Following Cardiac Surgery: Focus on Ultrasonographic Assessment. J Ultrasound Med 2023;42:2481-90. [Crossref] [PubMed]
3. Slutsky AS, Ranieri VM. Ventilator-induced lung injury. N Engl J Med 2013;369:2126-36. [Crossref] [PubMed]
4. Brochard L, Slutsky A, Pesenti A. Mechanical Ventilation to Minimize Progression of Lung Injury in Acute Respiratory Failure. Am J Respir Crit Care Med 2017;195:438-42. [Crossref] [PubMed]
5. Vassilakopoulos T, Petrof BJ. Ventilator-induced diaphragmatic dysfunction. Am J Respir Crit Care Med 2004;169:336-41. [Crossref] [PubMed]
6. Bruni A, Garofalo E, Pasin L, et al. Diaphragmatic Dysfunction After Elective Cardiac Surgery: A Prospective Observational Study. J Cardiothorac Vasc Anesth 2020;34:3336-44. [Crossref] [PubMed]
7. Laghlam D, Lê MP, Srour A, et al. Diaphragm Dysfunction After Cardiac Surgery: Reappraisal. J Cardiothorac Vasc Anesth 2021;35:3241-7. [Crossref] [PubMed]
8. Laghlam D, Naudin C, Srour A, et al. Persistent diaphragm dysfunction after cardiac surgery is associated with adverse respiratory outcomes: a prospective observational ultrasound study. Can J Anaesth 2023;70:228-36. [Crossref] [PubMed]
9. Jimenez-Santana JD, Azzolina D, Scaramuzzo G, et al. Ultrasound-detected postoperative diaphragmatic dysfunction and its association with pulmonary complications.A systematic review and meta-analysis. J Clin Anesth 2025;107:112027.
10. Colantuono S, Menzella F, Mari PV, et al. Patient response and remission in respiratory disease: Special focus on severe asthma and chronic obstructive pulmonary disease. J Int Med Res 2025;53:3000605251340894. [Crossref] [PubMed]
11. Mari PV, Carriera L, Ielo S, et al. Beyond Triple Therapy: Anti-Type 2 Biologic Therapies for Eosinophilic COPD and Respiratory Comorbidities - a Review of Clinical Evidence. Curr Allergy Asthma Rep 2026;26:29. [Crossref] [PubMed]
12. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. [Crossref] [PubMed]
13. Zhu X, Wu D, Zhang H, et al. Imaging and clinical features of diaphragm dysfunction after cardiac sternotomy. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue 2025;37:657-63. [Crossref] [PubMed]
14. Tralhão A, Cavaleiro P, Arrigo M, et al. Early changes in diaphragmatic function evaluated using ultrasound in cardiac surgery patients: a cohort study. J Clin Monit Comput 2020;34:559-66. [Crossref] [PubMed]
15. Maranta F, Cianfanelli L, Rizza V, et al. Diaphragm Dysfunction after Cardiac Surgery: Insights from Ultrasound Imaging during Cardiac Rehabilitation. Ultrasound Med Biol 2022;48:1179-89. [Crossref] [PubMed]
16. Huai H, Ge M, Zhao Z, et al. Early diaphragm dysfunction assessed by ultrasonography after cardiac surgery: a retrospective cohort study. Front Cardiovasc Med 2024;11:1457412. [Crossref] [PubMed]
17. Singh SK, Tagliafierro M, Kirilina D, et al. Minimally Invasive Approach for Mitral Valve Surgery Is Associated With Increased Risk of Postoperative Diaphragm Dysfunction. Innovations (Phila) 2026;21:144-9. [Crossref] [PubMed]