The results of this network meta-analysis based on the Hasse diagram and P-scores suggests ASV and TPNS as the most effective treatment for CSA in patients with HFrEF, in terms of their effects on AHI and ESS. This conclusion is based on ranking from P-scores and Hasse diagrams which are based solely on effect size and SEs. It is also noteworthy to mention that in AHI reduction all treatments were superior to GDMT and in the overall analysis no inter-treatment differences were found. This analysis, however, showed high heterogeneity and upon exploring this heterogeneity with sensitivity analysis, two studies (22, 27) were found to be contributing to the overall heterogeneity. The sensitivity analysis showed a statistically significant difference between ASV and all others. In this sensitivity analysis, the study by O’Connor et al[27] stood out as one of the outliers and was excluded. This study had low statistical power primarily because after the results of the SERV-HF trial[12] became available further recruitment of participants was stopped. The other outlier was the study by Fietze et al[22] and while the study was similar to most other PAP intervention studies, in that the participants showed significant reductions in sleep-disordered breathing with PAP intervention along with an improvement in LVEF, the only difference between this study and others was that it reported respiratory disturbance index as opposed to AHI (which was reported in all the others) and this was defined by the authors as the sum of the Cheyne-Stokes Apnea index, periodic breathing index, obstructive apnea index and mixed apnea index. With the sensitivity analysis, although the ranking of studies did not change, the differences between ASV and TPNS as well as its difference with BPAP-BUR and CPAP did become significant. It is important to note here that TPNS unlike its other active comparators (in this network meta-analysis) does not treat any obstructive apneas or obstructive hypopneas.[29] Despite this limitation of TPNS, it is remarkable that in the overall AHI reduction, TPNS out-performed BPAP-BUR and all other interventions in ESS improvement, actually meeting the minimal clinically important difference threshold with GDMT (based on both direct and indirect estimates) and also when compared with ASV, BPAP-BUR and CPAP (based on indirect estimates). It was not possible to separately analyze the data on central apnea index in addition to the overall AHI, and this is because not many studies included in this network meta-analysis reported this index separately.
This network meta-analysis evaluated the comparative efficacy of only four selected treatment comparisons and while it shows superior efficacy of ASV in AHI reduction in patients with HFrEF with CSA with LVEF < 50%, use of ASV especially in those with LVEF < 45% is currently contraindicated due to the findings of SERV-HF trial that showed an absolute annual risk of cardiovascular death of 10% in patients with HFrEF-CSA who were randomized to ASV vs. 7.5% in those who were in the control group.[12] The results of the much-awaited ADVENT-HF trial may provide more clarity on future use of ASV in this particular population.[30] Two meta-analyses have shown conflicting results on the effects of ASV on LVEF- one showing that while ASV significantly reduced AHI, the effects on LVEF improvement were not significant[31] and the other showing significant improvement in LVEF with ASV as compared to control.[32] This network meta-analysis did not evaluate the comparative efficacy of treatments on LVEF improvement primarily because no such data was available from the included study on TPNS[21] and hence impossible to form a ‘network’ with other interventions in a network meta-analysis. Additionally, given the different follow-up durations of the various interventions in this network meta-analysis, a 6-month data of LVEF changes from TPNS would conceivably not have been a fair assessment when compared to others in the network meta-analysis, primarily because a period of approximately 3 months is needed to optimally titrate nerve ‘stimulation’[29] and therefore one would not expect LV remodeling to be evident until after 6 months of maximum active therapy or in other words after 9–12 months of randomization. Furthermore, while ASV, BPAP-BUR and CPAP interventions in this network meta-analysis treated both central and obstructive apneas and hypopneas, given that TPNS only addresses central apneas, an assessment of LVEF function in a network meta-analysis would not be reasonable. Even so, while the ‘remedē System’ pivotal trial[21] did not report the 6-month data on LVEF changes, it is reassuring to know that the 12-month [29] and 36-month [33] data did show small but significant improvements in LVEF compared to baseline. Furthermore, while there are issues of some sort with other interventions, like device safety issues with ASV in CSA patients with low LVEF[12], and in general poor adherence with any PAP (CPAP or BPAP), the overall safety profile of TPNS as shown in the 36 months follow-up is quite promising.[33] While the debate continues- whether CSA-CSR is a friend or foe in HF patients and that perhaps the hyperventilation during CSR may actually be protective in these patients[34-36], one theory on the possible physiological links between ASV and the noted cardiovascular adverse outcomes in SERV-HF trial, that has gained much attention relates to the possible effects of high PAP on intrathoracic pressure and downstream effects. Accordingly, high PAP leads to high intrathoracic pressure, in turn leading to decrease in venous return, decreasing thereby the right ventricular stroke volume and consequently decreasing LV filling and cardiac output and eventually tipping the sympatho-vagal balance in these HF patients to the sympathetic side, which can be deleterious to HF patients.[37] A recent proof-of-concept study showed that neither nocturnal CPAP nor ASV favorably altered the sympathetic tone at night in CSA patients with systolic HF and patients on ASV had a significantly lower CI as compared to those on CPAP.[38] Not only that, but another recent study [39] showed that use of ASV in these patients significantly decreased N3 sleep (during which vagal tone predominates) and significantly increased REM (during which sympathetic nervous activity is known to be highest or close to wakefulness activity) and N2 sleep (during which burst of sympathetic activity have been known to occur). In this context, the mechanism of action of phrenic nerve stimulation is different and opposite to that of ASV, in that while ASV increases intrathoracic pressure by PAP, TPNS via neurostimulation triggers normal breathing via diaphragmatic contraction and thus generates negative intrathoracic pressure, thereby favoring venous return to the heart.[29]
There are several limitations to this network meta-analysis. This network meta-analysis was limited in that it could not evaluate other outcomes of interest such quality-of-life metrics, or changes in LVEF, or mortality with each intervention. However as explained earlier, without having all included studies report data in similar units, one cannot create a ‘network’ in a network meta-analysis to analyze comparative efficacies or effectiveness of interventions specifically with regards to quality-of-life outcomes and LVEF. Analysis of mortality or other long term adverse outcomes was outside the scope of this network meta-analysis which primarily aimed at analyzing the efficacy of interventions and included studies with follow up data not long enough to be able to compute any meaningful data on such outcomes. Furthermore, the control groups in the included RCTs, which were used as common comparator or reference in the network meta-analysis, were regarded as ‘GDMT’, however, it is possible that the medical therapy may have varied in its composition in different studies over time as the guidelines evolved. Lastly, and most importantly a strict inclusion criteria in this network meta-analysis limited the number of studies in each intervention category and the efficacy data on TPNS is derived from only one RCT.[21] Nevertheless, this is the first network meta-analysis to our knowledge assessing the comparative efficacy of all interventions in CSA patients with HFrEF.
In conclusion, this network meta-analysis suggests that both TPNS and ASV are superior to other interventions in managing sleep-disordered breathing in CSA patients with HFrEF and that TPNS therapy as opposed to other interventions can also improve subjective daytime sleepiness in these patients. This network meta-analysis may have to be updated when the results of ADVENT-HF trial[30] are published.