COVID-related dysphonia and persistent long-COVID voice sequelae: A systematic review and meta-analysis

The study examines how frequently dysphonia occurs in COVID-19 and whether it persists after recovery, addressing an under-recognized symptom that impacts communication, self-confidence, and quality of life. While COVID-19 is known to affect the aerodigestive tract with symptoms such as cough, dyspnea, sore throat, and chemosensory dysfunction, the exact prevalence and risk factors for voice impairment have not been synthesized. With long-COVID defined by WHO as symptoms lasting at least two months and occurring three months from onset, it remains unclear how dysphonia prevalence changes after recovery and which factors influence persistence. The objective was to determine global prevalence during acute infection and post-recovery, and to explore epidemiological risk factors, to inform management and prevention strategies.

This systematic review and meta-analysis adhered to PRISMA guidelines and was registered in PROSPERO (CRD42022383399). Databases searched included PubMed, Embase, ScienceDirect, Cochrane Library, and Web of Science on December 15, 2022, covering literature from December 2019 to December 2022 without language restriction; references of included studies were also screened. Inclusion criteria: prospective/retrospective cohort, case-control, cross-sectional studies, or clinical trials; availability of abstract and full-text; clear definition/diagnosis of dysphonia; sufficient extractable data on dysphonia prevalence. Exclusions: age <18 years; populations mainly of occupational/frequent voice users; studies not excluding pre-existing dysphonia/voice complications prior to COVID-19. Data extracted included study characteristics, demographics (age, sex, smoking), voice-related comorbidities, assessment modalities (objective: endoscopy/acoustic analysis; subjective: clinician judgement/self-report/questionnaires), tracheostomy/intubation rates, prevalence during and after infection, and follow-up durations. Risk of bias was assessed using the STROBE checklist by two reviewers. Statistical analysis used random-effects models to pool prevalence with 95% CIs; heterogeneity was quantified with I2. Meta-regressions evaluated mean age, gender, tracheostomy/intubation percentage, and assessment modality (objective with/without subjective vs subjective only). Sensitivity analysis applied one-study removal. Publication bias was evaluated via funnel plots and Egger’s tests. Two timeframes were analyzed: during acute infection, and post COVID-19 (after clinical recovery, typically ≥1 month), with a subgroup for long-COVID (≥3 months after recovery). Analyses were conducted in Comprehensive Meta-Analysis v3.

• 21 studies, 13,948 patients included.
• During acute COVID-19 (13 studies; 12,913 patients): weighted dysphonia prevalence 25.1% (95% CI: 14.9–39.0%); high heterogeneity (I2 = 97.1%, P < .001). Male sex associated with lower prevalence (coefficient −0.116, 95% CI −0.196 to −0.036; P = .004). Mean age, tracheostomy/intubation rate, and assessment modality were not significant. Sensitivity analyses did not change results. Egger’s test not significant (P = .889).
• Post COVID-19 (14 studies; 1,716 patients): weighted prevalence of persistent dysphonia 17.1% (95% CI: 11.0–25.8%); high heterogeneity (I2 = 92.9%, P < .001). No significant associations with mean age, gender, tracheostomy/intubation rate, or assessment modality. Egger’s test not significant (P = .072).
• Long-COVID subgroup (≥3 months; 933 patients): 20.1% (95% CI: 8.6–40.2%) had persistent dysphonia; high heterogeneity (I2 = 95.8%).
• Approximately 70% of those with dysphonia during acute infection continued to experience dysphonia after recovery.

Findings show dysphonia affects about one-quarter of patients during acute COVID-19 and persists in a substantial proportion after recovery, underscoring its clinical importance. Biological plausibility includes SARS-CoV-2 entry via ACE2 and TMPRSS2, which are highly expressed in the upper aerodigestive tract, particularly the vocal tract, leading to inflammatory laryngeal manifestations (e.g., vocal fold immobility, glottic/subglottic stenosis). The higher prevalence in females during acute infection may reflect sex-related immune responses rather than differential ACE2/TMPRSS2 expression, aligning with literature indicating more aggressive immune responses in females. Tracheostomy/intubation did not significantly influence prevalence when analyzed with other covariates, consistent with evidence that post-extubation voice symptoms are often transient and may not drive mid- to long-term dysphonia. Subjective assessment methods yielded comparable prevalence estimates to objective or combined methods, a practical consideration during pandemic-related restrictions. Persistent dysphonia post-recovery did not correlate with analyzed covariates, suggesting factors driving long-term symptoms may differ from acute-phase determinants or be multifactorial. Clinically, persistent voice sequelae may benefit from targeted therapies (e.g., corticosteroids, botulinum toxin where indicated), surgical intervention for structural lesions, and voice rehabilitation through speech-language therapy.

About 25.1% of COVID-19 patients, with higher risk among females, experience dysphonia during acute infection. Following recovery, 17.1% have persistent dysphonia, and in long-COVID cohorts, prevalence is approximately 20.1%, implying roughly 70% persistence among those initially dysphonic. Clinicians should proactively evaluate and manage voice symptoms during and after COVID-19. Future research should address severity, modifiable risk factors, treatment effectiveness, and variant-specific impacts in the post-pandemic era.

• Severity of dysphonia was not assessed.
• Potential influencers such as smoking status and minor comorbidities were not meta-analyzed due to insufficient data.
• Treatments and rehabilitation outcomes were not analyzed; adherence and loss to follow-up could bias such analyses.
• High between-study heterogeneity; some risk of bias related to population size determination, study-specific bias, and participant selection.
• Results may not fully reflect prevalence in the Omicron-dominant period; study quality varied across included observational studies.