Swallowing therapy for dysphagia in acute and subacute stroke

Post-stroke dysphagia ((PSD) impaired swallowing following stroke), increases risk of death, yet the effects of potential treatments are unclear. This review updates the 2018 review.

To assess the effects of swallowing therapies on swallowing impairment (SI) among stroke survivors with dysphagia, within six months of stroke onset.

We searched MEDLINE, Embase, OVID, three other databases and trials registries on 31 July 2025.

Randomised controlled trials (RCTs) that compared interventions for PSD with usual care.

Critical outcome: SI. Important outcomes: dysphagia at end of trial (DET), penetration/aspiration scale (PAS) score, chest infection or pneumonia (CI/P), and decannulation.

We assessed risk of bias using the Cochrane Risk of bias 1 tool (RoB 1).

We calculated odds ratios (ORs), mean differences (MDs) and standardised mean differences (SMDs), with 95% confidence intervals (CIs), using random-effects models. We assessed the certainty of evidence with GRADE.

We included 181 trials with 11,500 participants (140 new trials (8840 participants) in this update) that compared usual care with acupuncture, behavioural interventions, drug interventions, neuromuscular electrical stimulation (NMES), pharyngeal electrical stimulation (PES), physical stimulation, theta burst stimulation (TBS), transcranial direct current stimulation (tDCS), transcranial magnetic stimulation (TMS), vagal nerve stimulation and combined interventions.

Acupuncture may reduce SI but effects are uncertain (SMD -1.20, 95% CI -1.50 to -0.90; 37 RCTs, 3335 participants; very low-certainty evidence); probably reduces DET (OR 0.40, 95% CI 0.33 to 0.47; 36 RCTs, 3647 participants) and PAS (MD -0.90, 95% CI -1.25 to -0.56; 8 RCTs, 496 participants) (both moderate-certainty evidence); and may reduce CI/P (OR 0.47, 95% CI 0.25 to 0.85; 4 RCTs, 426 participants; low-certainty evidence). Behavioural interventions may reduce SI (SMD -1.17, 95% CI -1.60 to -0.74; 26 RCTs, 1224 participants) and DET (OR 0.41, 95% CI 0.30 to 0.57; 9 RCTs, 958 participants), and have little to no effect on decannulation (OR 0.17, 95% CI 0.01 to 3.88, 1 RCT, 22 participants), but effects are uncertain (all very low-certainty evidence); and may reduce PAS (MD -0.99, 95% CI -1.42 to -0.55; 16 RCTs, 586 participants) and CI/P (OR 0.45, 95% 0.27 to 0.76; 8 RCTs, 784 participants) (both low-certainty evidence). Drug interventions may reduce SI but effects are uncertain (SMD -0.89, 95% CI -1.62 to -0.15; 3 RCTs, 140 participants; very low-certainty evidence); may have little to no effect on DET (OR 0.48, 95% CI 0.07 to 3.35; 1 RCT, 17 participants; low-certainty evidence); and probably reduce CI/P (drug therapy: 27% versus control: 87%, OR 0.06, 95% CI 0.01 to 0.21; 1 RCT, 60 participants; moderate-certainty evidence). NMES probably reduces SI (SMD -0.69, 95% CI -1.01 to -0.37; 20 RCTs, 866 participants; moderate-certainty evidence); and may reduce DET (OR 0.43, 95% CI 0.28, 0.67; 7 RCTs, 383 participants) but not PAS (MD -0.71, 95% CI -1.75 to 0.34; 6 RCTs, 155 participants) or CI/P (OR 0.30, 95% CI 0.03 to 3.03; 4 RCTs, 189 participants) though the effects are uncertain (all very low-certainty evidence). PES may have little to no effect on SI (SMD -0.15, 95% CI -0.46 to 0.16; 5 RCTs, 272 participants), but may reduce PAS (MD -0.74, 95% CI -1.44 to -0.03; 5 RCTs, 195 participants) and CI/P (OR 0.33, 95% CI 0.12 to 0.93; 2 RCTs, 88 participants); all low-certainty evidence. PES probably has little to no effect on DET (DET: 68% control: 78%; OR 0.55, 95% CI 0.15 to 2.11; 3 RCTs, 66 participants) but probably increases decannulation (OR 0.09, 95% CI 95% CI 0.03 to 0.28; 2 RCTs, 98 participants; moderate-certainty evidence). Physical stimulation may reduce SI (SMD -0.78, 95% CI -1.29 to -0.26; 3 RCTs, 228 participants), DET (OR 0.29, 95% CI 0.14 to 0.60; 4 RCTs, 350 participants), and CI/P (OR 0.20, 95% CI 0.05 to 0.76; 1 RCT, 96 participants); all low-certainty evidence. TBS may reduce SI (SMD -0.81, 95% CI -1.06 to -0.56; 5 RCTs, 280 participants; low-certainty evidence) and PAS but effects are uncertain (MD -1.26, 95% CI -2.04 to -0.48; 5 RCTs, 280 participants; very low-certainty evidence). tDCS: probably reduces SI (SMD -0.63, 95% CI -0.81 to -0.44; 14 RCTs, 496 participants; moderate-certainty evidence), may reduce DET (OR 0.28, 95% CI 0.09 to 0.88; 3 RCTs, 118 participants; low-certainty evidence), but effects on PAS are unclear (MD -0.82, 95% CI -1.83 to 0.18; 4 RCTs, 135 participants; very low-certainty evidence). TMS: may reduce SI (SMD -1.03, 95% CI -1.38 to -0.69; 17 RCTs, 661 participants; low-certainty evidence) and likely reduces PAS (MD -1.19, 95% CI -1.53 to -0.85; 12 RCTs, 519 participants; moderate-certainty evidence). Vagus nerve stimulation probably reduces SI (SMD -1.27, 95% CI -1.97 to -0.58; 1 RCT, 39 participants; moderate-certainty evidence). Combined therapies probably reduce SI (SMD -1.45, 95% CI -2.04 to -0.85; 8 RCTs, 334 participants), but not CI/P (OR 0.08, 95% CI 0.00 to 1.47; 1 RCT, 70 participants) (both moderate-certainty evidence), and may make little to no difference to DET (OR 0.19, 95% CI 0.03 to 1.34; 2 RCTs, 65 participants; low-certainty evidence).

Conclusions from our analyses are limited by small trials at high or unclear risk of bias, publication bias and inconsistent results. Large high-quality trials on specific interventions are needed.

No dedicated funding.

Protocol (1997), DOI not available Original review (1999) DOI 10.1002/14651858.CD000323 First update (2012) DOI 10.1002/14651858.CD000323.pub2 Second update (2018) DOI 10.1002/14651858.CD000323.pub3.