Eight crossover trials were included in the review (n=129 patients, range 15 to 20). All trials received Jadad scores of either 1 or 2 (scale 0 to 5), which indicated poor study quality. Validity assessment by individual criteria were not reported.
Overall, there was a statistically significant mean difference in apnea hypopnoea index outcome (-5.40, 95% CI -9.82 to -0.98) that favoured overdrive pacing (based on eight trials), but no statistically significant mean difference between treatments that used the minimum SaO2 measure (based on seven trials).
There was significant heterogeneity for the apnea hypopnoea index outcome (I2=67.7%). This disappeared when the eight trials were stratified into two subgroups according to the proportion of patients with obstructive rather than central sleep apnoea syndrome. For the two trials (n=31 patients) where most of the patients had central sleep apnoea syndrome, apnea hypopnoea index outcomes favoured overdrive pacing far more substantially (MD -17.08, 95% CI -23.25 to -10.91) than for the six trials (n=98 patients) where most patients had obstructive sleep apnoea syndrome (MD -2.94, 95% CI -5.33 to -0.54). In one trial where most patients had central sleep apnoea syndrome and minimum SaO2 was assessed, outcomes favoured overdrive pacing (MD 4.00, 95% CI 2.48 to 5.52) over nonpacing; this was not the case for the six trials where most patients had obstructive sleep apnoea syndrome.
Four of eight correlations for apnea hypopnoea index pooled analysis and four of the seven correlations for the minimum SaO2 pooled analysis were imputed as trial data did not allow them to be estimated. Sensitivity analysis performed to determine whether the assumptions made in these imputations affected the direction and statistical significance of either pooled outcome judged this not to be the case.