Treatment of chronic respiratory failure with long-term home noninvasive ventilation (NIV) to reduce symptom load and improve survival in patients with slowly progressive neuromuscular and chest wall disorders (NMD, CWD) is part of standard clinical practice.1,2 Guideline criteria for NIV initiation in stable restrictive lung disease is targeted toward relief of symptoms, which is achieved by treating nocturnal hypoventilation and chronic respiratory hypercapnia with home NIV.2–5 Targeted reduction of carbon dioxide is one of the main goals of NIV.6 This implies a combination of sufficient ventilatory support to treat alveolar hypoventilation, and sufficient usage of this treatment.

In chronic obstructive pulmonary disease (COPD), a common indication for home NIV, the concept of “high-intensity” NIV has led to a paradigm shift over the past decade and is gaining growing consideration.7–11 High-intensity NIV can be defined as a strategy that consists of adjusting ventilator parameters with the specific goal of reducing transcutaneous carbon dioxide (TcCO2) and arterial partial pressure of carbon dioxide (PaCO2) by setting sufficient levels of pressure support (PS) (or tidal volume, VT) and back-up respiratory rate (BURR).12–15 Current clinical practice guidelines clearly state that NIV should be set with the aim of reducing or normalising PaCO2 levels in individuals with COPD.10,11

However, data suggest that NIV may not always effectively improve PaCO2 levels and reverse hypoventilation-related symptoms.16–18 For instance, 12–40% of individuals with NMD have residual hypercapnia under NIV; furthermore, residual hypercapnia is associated with negative outcomes.19 Despite these findings, the concept of using higher levels of NIV intensity in individuals with NMD and CWD has never really been discussed and deserves to be evaluated.20

Our aim was to determine the extent to which, during the initial period of NIV initiation, the intensity of NIV parameters influences PaCO2 levels in individuals with chronic respiratory failure due to slowly progressive NMD or CWD. Given the small number of studies that have addressed this question, and given the heterogeneity of published data regarding participants’ diagnoses and ventilatory modes used, we undertook a systematic review with meta-analysis of individual participant data (IPD).

This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic Review and Meta-Analyses of Individual Participant Data (PRISMA-IPD) statement.21 The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42021245121: www.crd.york.ac.uk/prospero/).

Data synthesis and analysis

All analyses were conducted according to the predefined statistical analysis plan outlined in the protocol. Given the low rate of missing data, we performed a complete case analysis. One outlier with a baseline PaCO2 level>15kPa was excluded from the analysis of PaCO2 and two outliers with a PaO2 level>16kPa at study endpoints were excluded from the analysis of PaO2.

Diurnal PaCO2 level at study endpoints was analysed using a generalised linear mixed model with a random intercept for study. In the base model, baseline PaCO2, daily NIV usage at study endpoints, and disease category (NMD versus CWD) were included as fixed effects. Then, all other variables were tested one by one and included in the final model if p<0.2 and the rate of missing data was<10%. Lastly, the impact of ventilatory parameters (PS, VT and BURR) was tested. We defined NIV intensity as the product of PS and BURR for pressure-cycled modes, and the product of VT and BURR for volume-cycled and hybrid modes. We used the Z-scores of these values to obtain a unified measure of NIV intensity regardless of the NIV mode. By definition, mean NIV intensity of the whole cohort corresponds to a Z-score of 0. We defined the mean – 1SD, mean, mean+1SD, and mean+2SD of NIV intensity Z-scores as low, medium, high and very high NIV intensity. A conversion chart between ventilatory parameters and NIV intensity Z-scores is presented in Fig. 1. Interactions between NIV intensity, disease category and baseline PaCO2 level were also tested.

Fig. 1.

Values of NIV intensity Z-scores according to backup respiratory rate for pressure-cycled modes (panel A), and for volume-cycled and hybrid modes (panel B). Noninvasive ventilation (NIV) intensity was defined as the product of pressure support (PS) and backup respiratory rate (BURR) for pressure-cycled modes, and of tidal volume (VT) and BURR for volume-cycled and hybrid modes. Z-scores represent the number of standard deviations above or below the mean NIV intensity used in the included studies. For pressure-cycled modes (panel A), the combination of a BURR set at 14breaths/min and a PS set at 10cmH2O corresponds to a Z-score of 0 (medium NIV intensity). The combination of a BURR set at 14breaths/min and a PS set at 20cmH2O corresponds to a Z-score of 2 (very high NIV intensity). For volume-cycled or hybrid modes (panel B), the combination of a BURR set at 18breaths/min and a VT set at 5mLkgPBW−1 corresponds to Z-score of −1 (low NIV intensity). The combination of a BURR set at 18breaths/min and a VT set at 15mLkgPBW−1 corresponds to a Z-score of 1 (high NIV intensity). BURR, backup respiratory rate; NIV, noninvasive ventilation; PS, pressure support.

(0.15MB).

For secondary outcomes (PaO2 and daily NIV usage), the same approach was used for the construction of the final model. Finally, to investigate the relation between interface type and NIV intensity, we identified NIV intensity as the dependent variable, and we included interface type as a fixed effect in the base model, along with baseline PaCO2 level, disease category and daily NIV usage.

In the final models, the following assumptions were verified: linearity, absence of collinearity in the predictors, homoscedasticity, normality of residuals, absence of influential data points and independence. All statistical analyses were performed with R and Jamovi (Gamjl package), and R packages lmer4 and lmerTest.

The primary aim of this systematic review with meta-analysis of IPD was to evaluate the effect of NIV intensity on PaCO2 level in individuals with slowly progressive NMD or CWD initiated on long-term home NIV. We used an original method involving the Z-score principle, which allowed us, based on the included studies, to define the intensity of NIV regardless of the ventilatory mode. In the whole study sample, NIV intensity was not significantly associated with PaCO2 level at study endpoints. However, the effects of NIV intensity differed between the underlying disease categories: no significant effect of NIV intensity was found for NMD, whereas in CWD with the most severe baseline hypercapnia, higher NIV intensities were associated with greater reductions in PaCO2 levels. Neither daily NIV usage nor interface type were associated with NIV intensity.

The severity of the respiratory impairment at the time of NIV initiation could explain the discrepancy in the effect of NIV intensity on PaCO2 level at study endpoints between participants with NMD and CWD. Participants with NMD had moderate impairment of respiratory function and gas exchange at baseline. It is therefore conceivable that they had no major impairment of thoraco-pulmonary compliance or neural respiratory drive. Moreover, patients with NMD are often closely monitored in specialised centres; therefore, respiratory failure is managed early. Participants may have had other symptoms for which the use of NIV was indicated, such as nocturnal hypercapnia or hypoxemia, orthopnoea or reduced muscle strength.2 Consequently, the change in PaCO2 level was achieved even with low intensity NIV. These results contrast with those of a recent retrospective study by our group that suggested that higher levels of ventilatory support were associated with lower nocturnal TcCO2 in individuals with NMD.20 However, the participants in that study had been treated with NIV for more than 8 years on average and therefore likely had more advanced disease. Additionally, the analysis was based on nocturnal, rather than diurnal evaluation of PCO2, which may also explain the stronger relationship between NIV intensity and PCO2.

In the present meta-analysis, respiratory function and gas exchange at baseline were more impaired in the individuals with CWD than in those with NMD. In CWD, increasing NIV intensity could compensate for reduced compliance, which is a major determinant of alveolar hypoventilation in these disorders.32,33 Although the model showed that NIV intensity was associated with a greater improvement in hypoventilation in the individuals with CWD with the highest levels of hypercapnia, a paradoxical effect of NIV intensity might occur in those with moderate or no hypercapnia; in this situation, the benefits of NIV intensity may be limited.

It is noteworthy that the sensitivity analysis including only patients with diurnal hypercapnia (PaCO2≥6kPa) at treatment initiation did not alter the direction of the results of our main analysis, thereby supporting the validity of our findings regardless of baseline PaCO2.

Our results also support evidence that the amount of daily NIV usage is an essential determinant of improvement in daytime PaCO2 level.34–37 This finding highlights the importance of encouraging good adherence to treatment to obtain sufficient daily usage, at least during the first months, rather than immediately increasing NIV intensity to reduce PaCO2 levels. Close follow-up, for instance by telemonitoring or with specific procedures such as polysomnography-directed titration, could be useful to increase daily NIV usage in the initial phase of treatment.29,38

Finally, our results did not show any association between NIV intensity and the type of mask used (nasal vs. oronasal) in either NMD or CWD. This contrasts with the results of a recent meta-analysis in individuals with COPD or OHS that showed that the inspiratory positive airway pressure level tended to be higher (1.42 [−0.04, 2.88]cmH2O) in individuals fitted with an oronasal mask.39

This meta-analysis has several limitations. First, among the 16 studies that fulfilled our inclusion criteria, only 7 authors shared their IPD. The studies for which we did not obtain IPD were older, and alteration of FVC at the time of NIV initiation was significantly more pronounced, which likely reflects the current trend to introduce NIV at earlier stages of disease progression.2,3 Although these considerations may have induced a selection bias, it should be noted that very few studies restricted their inclusion criteria to our population of interest. Consequently, our research question could not be addressed with aggregated data, which made it necessary to obtain IPD. The resulting sample size was therefore limited and, even though this meta-analysis gathered one of the largest datasets analysed in such a population,16 it may have been underpowered and so the results must be interpreted cautiously, particularly the modelling of the highest NIV intensity levels for which the confidence intervals were quite large.

Second, we combined data from prospective cohort studies and RCTs, which could have led to heterogeneity in the results. However, in the multivariate models, study design was not significantly associated with any of the outcomes of interest. In addition, in the RCTs, we only used data from the individuals who received treatment, and no control groups were available in the cohort studies. Therefore, we cannot differentiate between the effect of the natural course of the diseases and the effect of NIV. Although we tried to minimize this bias by testing and adjusting for participants’ baseline characteristics (especially baseline PaCO2), we cannot exclude the presence of residual and unmeasured confounding.

Third, the data provided did not allow us to conclude on the effect of NIV intensity on nocturnal variables, especially nocturnal hypoventilation, which is a very common and important indication for ventilation in these disorders.2 The effect of NIV intensity on nocturnal symptoms should be evaluated in clinical trials, particularly in this population in which improving sleep quality (as well as health-related quality-of-life) is a major target of long-term home NIV.6

Finally, the NIV parameters by which we defined “high” levels of NIV intensity are notably lower than what is encountered in other aetiologies such as COPD.12–15,40 Nevertheless, these values reflect what is actually documented in the literature in this heterogeneous population, and should be further investigated.

This meta-analysis found no significant effects of NIV intensity on PaCO2 levels in individuals with NMD or CWD initiated on long-term home NIV. More specifically, the effects of NIV intensity at the time of treatment initiation are not uniform across populations; higher NIV intensities may be of benefit to individuals with CWD and the most severe levels of baseline hypercapnia. Importantly, the amount of daily NIV usage, whatever the settings, appears to be a decisive independent factor in determining NIV effectiveness on gas exchange. During the early period of NIV implementation, achieving sufficient therapy usage seems to be a consideration that likely should prevail over the intensity of settings to increase alveolar ventilation. Further prospective studies should be conducted to confirm these results.

MD, AL, ML, CK and J-CB contributed to the conception and design of the study. All the authors have written or edited the manuscript.

AH, MP, AHN, LMH, MB, AY, MEH, NH and PJW provided complete IPD from their respective studies. Each co-author made substantial contributions to the manuscript; drafted sections of the manuscript and revised it critically for important intellectual content; provided final approval of the version to be published; agreed to be accountable for all aspects of the manuscript and to ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Not required.

No data are available. All the de-identified individual participant data collected in this systematic review and meta-analysis must be requested from each author individually.

The authors do not declare a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

MD reports personal fees from Air Liquide Medical Systems, Breas Medical AB, and ResMed SAS, outside the submitted work; AL reports consulting fees from Air Liquide Medical Systems, outside the submitted work; ML is a part time employee of Air Liquide Medical Systems, outside the submitted work; AH reports grants from ZonMw VIMP, outside the submitted work; WW reports grants and personal fees from Löwenstein Medical, Germany, grants from Philips Respironics, USA, and personal fees from Sentec, Switzerland, outside the submitted work: LH and MEH report in-kind support from Philips Respironics to his research institute, outside the submitted work; PJW reports personal fees from Philips Respironics, outside the submitted work; HP reports personal fees from ASV Santé, SOS Oxygène, ISIS Medical, Breas Medical, ResMed, Sanofi – Genzyme, and Sanofi – Biogen, outside the submitted work; J-CB is employed by AGIR à dom (French home care provider), outside the submitted work. The other authors have no conflicts of interest to disclose.