Abstract

Objective: To systematically review the literature regarding the ability of clinical features to predict respiratory failure in patients with Guillain–Barré syndrome (GBS).

Data sources: We searched the PubMed and Ovid MEDLINE databases with the search terms “guillain barre syndrome” OR “acute inflammatory demyelinating polyneuropathy” OR “acute motor axonal neuropathy” OR “acute motor sensory axonal neuropathy” AND “respiratory failure” OR “mechanical ventilation”. We excluded articles that did not report the results of original research (eg, review articles, letters), were case reports or series (ten or fewer patients), were not available in English, reported research in paediatric populations (16 years of age or younger), or were interventional studies. Article quality was assessed with the Newcastle–Ottawa quality assessment scale.

Data synthesis: Thirty-four relevant studies were identified. Short time from symptom onset to hospital admission (less than 7 days), bulbar (odds ratio [OR], 9.0; 95% CI, 3.94–20.6; P < 0.001) or neck weakness (OR, 6.36; 95% CI, 2.32–17.5; P < 0.001), and severe muscle weakness at hospital admission were associated with increased risk of intubation. Facial weakness (OR, 3.74; 95% CI, 2.05–6.81; P < 0.001) and autonomic instability (OR, 6.40; 95% CI, 2.83–14.5; P < 0.001) were significantly more frequent in patients requiring intubation in our meta-analyses; however, the differences were not statistically significant in individual multivariable analysis studies. Four predictive models have been developed to assess the risk of respiratory failure for patients with GBS, each with good to excellent discriminative power (area under the receiver operating characteristic curve, 0.79–0.96).

Conclusions and relevance: Early identification of GBS patients at risk of respiratory failure could reduce the rates of adverse outcomes associated with delayed intubation. Algorithms that predict a patient’s risk of subsequent respiratory failure at hospital admission appear more reliable than individual clinical variables.