Twelve case series (n=342 patients) were included; five prospective and seven retrospective studies. Sample sizes ranged from eight to 101. Ten of the studies were classified as being subject to selection bias. There was no evidence of publication bias.
The pooled mean body mass index (10 studies) decreased from 55.3kg/m2 (95% CI 53.5 to 57.1) at baseline to 37.7kg/m2 (95% CI 36.6 to 38.9) after surgery. Mean apnoea-hypopnoea index (12 studies) decreased from 54.7 events per hour (95% CI 49.0 to 60.3) to 15.8 events per hour (95% CI 12.6 to 19.0). Heterogeneity was reported to be high (I2=74%), but it was unclear which pooling this applied to.
There was a trend towards lower apnoea-hypopnoea index (better outcome) post-surgery in the retrospective studies and in studies with selection bias, that is the poorer quality studies.
Individual patient data were available from six studies (n=80 patients). Patients classified as cured (apnoea-hypopnoea index less than 5) after surgery had a lower body mass index and were younger than those who had residual sleep apnoea. In logistic regression models, age and follow-up weight of less than 100kg independently predicted cure.