Post-extubation Pressures in Preterm Neonates: A CER Study

BACKGROUND Invasive mechanical ventilation (IMV) is a well-established risk factor for bronchopulmonary dysplasia and associated long-term complications. In recent years, a large body of research has evaluated and helped establish clinical practices aimed at avoiding initiation of IMV altogether by using non-invasive respiratory support modes such as CPAP in place of routine intubation and use of minimally invasive surfactant treatment. Despite such efforts, ~85% of preterm neonates <280 weeks' gestational age require IMV [Canadian Neonatal Network data - unpublished]. Furthermore, ~25% of preterm neonates extubated to non-invasive support ultimately require re-intubation [Canadian Neonatal Network data - unpublished], which itself is associated with complications including mortality and bronchopulmonary dysplasia. As such, further research on strategies to facilitate successful extubation is required. While there has been a considerable focus on comparing various modes of non-invasive support following extubation, there are limited published data on peri-extubation practices, particularly as they relate to the choice of post-extubation pressure levels.

In the meantime, there is considerable practice variability in choice of initial settings on both CPAP and NIPPV. A national-level survey (unpublished) demonstrated an almost even split - with half the sites choosing equal or lower pressures, whereas others choosing slightly higher pressures on the initial non-invasive respiratory support compared to pre-extubation pressures on IMV. As such, while the optimal post-extubation pressure level on non-invasive respiratory support that optimizes extubation success and clinical outcomes remains unknown, this natural practical variability affords a unique opportunity to conduct this comparative effectiveness research study using real-world data.

OBJECTIVE To determine the optimal initial non-invasive respiratory pressure support level following extubation in preterm neonates.

HYPOTHESIS The hypothesis is that higher pressures will lead to fewer re-intubations.

SPECIFIC RESEARCH QUESTION

Population: Preterm neonates GA <28 weeks' gestation, admitted to a participating centre who received any duration of mechanical ventilation.

Patient-level exclusion criteria:

Major chromosomal/genetic/congenital abnormalities Never received invasive mechanical ventilation (IMV) Received IMV, but never extubated to non-invasive respiratory support (NRS) Transferred to non-participating site while intubated Death prior to extubation Initial extubation was for withdrawal of care

Lack of any eligible extubation, defined as ALL of the following:

(i) ≥36 continuous hours on IMV; (ii) pre-extubation MAP ≤13 cmH2O; and (iii) on post-extubation NRS for ≥1 hour

Intervention group: "Higher Pressure"

Control groups: "Equal/Lower Pressure"

Primary outcome:

1. Re-intubation within 7 days Secondary outcomes:
2. Failure of initial NRS within 7 days (defined as any escalation beyond originally assigned settings, escalation to an alternate NRS mode, and/or re-intubation)
3. Presence of any one or more pre-defined re-intubation criteria within 7 days
4. Moderate-severe bronchopulmonary dysplasia
5. Death (any cause) prior to discharge
6. Death (any cause) prior to discharge OR moderate-severe bronchopulmonary dysplasia
7. Post-extubation days on any form of positive pressure respiratory support
8. Post-extubation days on invasive mechanical ventilation
9. New pulmonary air leak including pneumothorax, pulmonary interstitial emphysema, and/or pneumo-mediastinum within 7 days of extubation (per radiology report)
10. New gastro-intestinal perforation (any cause) within 7 days of extubation (per radiology report)
11. Necrotizing enterocolitis diagnosed any time post-extubation

METHODS

This will be a prospective comparative effectiveness research study based on real-world practices across participating tertiary NICUs in Canada. Centres have self-selected (or in cases where there is no preference, assigned):

1. Intervention ("Higher pressure") vs. control arm ("Equal/Lower pressure") pressure levels as the initial post-extubation pressure level, and
2. CPAP vs. NIPPV as the initial post-extubation mode

Note: Biphasic CPAP and Non-invasive high frequency ventilation (NIHFV) will be considered sub-forms of CPAP; whereas non-invasive neurally adjusted ventilatory assist (NIV-NAVA) will be considered a sub-form of NIPPV.

Only the first eligible extubation will be included for analyses; however, it will be recommended that centres follow their assigned strategy for all eligible extubations.

Peri-extubation respiratory management:

Pre-extubation: Among eligible babies (<28 weeks' GA) the first extubation from any ventilator mode that meets all of the following criteria will be included:

1. Extubation after ≥36 continuous hours of IMV
2. Measured pre-extubation Paw ≤13 cmH2O
3. NOT re-intubated within 1 hour of extubation

The decision and timing of extubation will rest completely with the medical team; specific extubation criteria will NOT be mandated. Data on actual pre-extubation modes and settings will be collected; the pre-extubation mean airway pressure (Paw) will be determined based on the measured value as provided by the ventilator prior to extubation.

Extubation procedure: The extubation procedure will be as per each individual centre's respective standard practice. Once the nasal interface is applied, the stipulated mode and NRS pressures should be in effect immediately, and the initial post-extubation mode and settings will be documented by the bedside healthcare provider.

Post-extubation: Any device capable of generating the stipulated pressures may be used, however only short binasal prongs and/or nasal mask interfaces will be permitted as the initial interface for the 1st 6 hours post-extubation.

Centres in the Higher NRS pressures arm will be required to maintain the stipulated NRS pressures as a minimum for at least 6 hours, a consensus-based choice of duration that will allow for sufficient stabilization of the lungs while avoiding premature weaning. Any wean within the first 6 hours will be considered a protocol deviation, and the reason for the wean will be documented. Pressure escalations will be permitted based on patient-need at clinicians' discretion, as well as escalation to alternate modes and/or intubation as needed. All other respiratory management will be as per the medical team.

Centres in the Equal/Lower NRS pressures arm will choose initial pressures as stipulated based on pre-extubation pressures. All subsequent management will be as per the medical team.

Re-intubation criteria: Suggested minimum re-intubation criteria will be provided if non-invasive support (despite escalation of pressure levels and/or use of rescue modes based on clinical team's discretion) is deemed insufficient. While these criteria will not be strictly enforced, patient records will be reviewed to identify patients who meet these criteria over 1st 7 days (one of the secondary outcomes).

Statistical Analyses:

The primary analyses will be conducted by first classifying patients based on treatment provided ("as-treated" classification). As such, each patient will be classified into Higher NRS pressures or Equal/Lower NRS pressures, regardless of the centre, based on the initial post-extubation strategy employed. Patients extubated to a strategy different than either of the two stipulated ones (e.g., extubation to Paw + 3 cmH2O, or extubation directly to room air) will be excluded from these analyses.

Demographics and baseline data will be summarized using appropriate univariate analyses - Chi-squared or Fisher's exact test for categorical variables and T-test or Wilcoxon Rank Sum test for continuous variables.

Centre characteristics will be descriptively summarized, grouped by study arm. Outcomes analyses - unadjusted: All outcomes will be summarized based on treatment provided and described using appropriate univariate analyses. Unadjusted odds ratios and unadjusted mean/median differences with 95% confidence intervals will be reported for categorical and continuous variables, respectively.

Outcomes analyses - adjusted: The investiagtors will conduct multivariable logistic and linear regression models for the primary and all secondary outcomes. These analyses will yield adjusted odds ratios or adjusted mean differences with 95% confidence intervals for categorical and continuous outcomes, respectively. Covariates for adjusted analyses - Adjusted modes for all outcomes will include the following three variables, regardless of the P values from univariate analyses, due to their biological relevance and importance to all outcomes: gestational age, sex, and the initial post-extubation NRS mode (i.e., CPAP or NIPPV). The following baseline and demographic variables will be considered for inclusion as covariates in the model for the primary outcome: antenatal steroids, hypertensive disorder of pregnancy, small for gestational age (<10%ile), multiple pregnancy, out-born status (i.e., delivered at a non-tertiary care centre), cesarian section, score for neonatal acute physiology (SNAP), multiple doses of surfactant, chronological age at extubation, administration of post-natal steroids at time of extubation, sepsis or meningitis or necrotizing enterocolitis any time prior to extubation, FiO2 pre-extubation, and pre-extubation Paw). These variables have been considered based on the available literature on clinical predictors of extubation failure and biological/clinical rationale. However, only those variables with a P value <0.20 (a conservative cut-off) based on univariate analyses will be included in the model. Furthermore, we will check for multi-collinearity amongst included variables to ensure avoidance of over-adjustment by monitoring the variance inflation factor, utilizing a cut-off value of <5.0. Finally, to account for centre-specific differences (i.e., clustering within centres), a generalized estimating equations approach will be used for all models with an appropriate covariance structure.

Note: The baseline/demographic variables under consideration as co-variates in models for the secondary outcomes will be similar to that for the primary outcome.

Sample Size:

Recent national-level data gathered for a previous similar comparative effectiveness research study contains data on re-intubation risk in neonates <28 weeks' gestation, the population of interest in this proposal. While this data does not account for post-extubation NRS settings, is it estimated that for patients extubated to Equal/Lower NRS pressures, the re-intubation risk will be 25%. In a recent (unpublished) survey, 86% of knowledge users across Canada responded that a 7.5% absolute risk reduction in re-intubation would be likely or very likely to change their clinical practice. To demonstrate this 7.5% absolute risk reduction (i.e., a re-intubation risk of 17.5% with use of Higher NRS pressures) will require a minimum of 367 patients in each arm (one-sided α of 0.05 and β of 0.20, from Pearson Chi-square test or Likelihood ratio test). As such, over a planned 3.5-year study recruitment period, allowing for up to 30% loss in eligibility (e.g., transfer to a non-participating centre while still on IMV, extubation for withdrawal of life-sustaining therapy, or extubation to an NRS mode/settings not part of the study), at least 525 patients will be recruited in both groups (i.e., total sample size 1,050). This will be the largest (and adequately powered) study to address this important question faced on a day-to-day basis by healthcare providers caring for preterm neonates.

Planned Sensitivity Analyses:

The investigators will conduct the following sensitivity analyses (for the primary outcome only):

1. Intention-to-treat analysis: In accordance with the EQUATOR network guidelines, patients will be classified to Higher vs. Equal/Lower NRS pressures based on the centre-assignment, irrespective of actual modes or pressures utilized. This will include patients who were extubated to a strategy not part of the study. Remaining analyses will be similar to as described earlier for the primary "as-treated" cohort.
2. Per-protocol analysis: Patients will be classified based on centre-assignment AND limited to only those infants who received the intended post-extubation NRS pressures. For example, from a centre that has self-selected to Higher NRS, only patients extubated to Higher NRS levels on CPAP or NIPPV will be included. Remaining analyses will be similar to as described earlier for the primary "as-treated" cohort.
3. Propensity-score matched analysis will be conducted to assess the accuracy of the primary analyses. Propensity-score matched analyses are considered more robust than traditional regression models in certain situations, especially when confounding by indication is suspected. In this model, each patient is assigned a probability (propensity-score) of receiving the assigned treatment (i.e., Higher vs. Equal/Lower NRS), based on the distribution of baseline and demographic variables that precede the treatment using a regression model. It is assumed that subjects with the same propensity score have a similar distribution of observed covariates, independent of the treatment received. We will confirm that the estimated propensity score balances covariates by regressing each covariate on treatment. For any unbalanced covariates, the model will be updated by allowing for effect modification (including covariate interactions) and non-linearity (including higher powers of continuous covariates) until all covariates are balanced. The propensity scores will then be used to match subjects with similar probabilities of receiving one of the two treatment strategies. Matching will be performed using the SAS macro "match.sas" (SAS Institute Inc., Cary, NC, USA), based on a caliper width of 0.2 times the standard deviation of the logit-transformed propensity score. Overall, this analysis is expected to create a cohort that approximates that of a traditional randomized controlled trial.

Planned Subgroup Analyses:

Subgroup analyses will be conducted for the following (for the primary outcome only):

1. Patients extubated to CPAP [i.e., comparison of Higher vs. Equal/Lower CPAP; excluding BiPhasic CPAP and NIHFV] and patients extubated to NIPPV [i.e., comparison of Higher vs. Equal/Lower NIPPV; excluding NIV-NAVA].
2. Chronological age at extubation ≤14 days [i.e., comparison of Higher vs. Equal/Lower NRS in babies extubated at age ≤14 days] and chronological age at extubation >14 days [i.e., comparison of Higher vs. Equal/Lower NRS in neonates extubated at age >14 days].
3. GA at birth <26 weeks [i.e., comparison of Higher vs Equal/Lower NRS in babies born at <26 weeks] and GA at birth ≥26 weeks [i.e., comparison of Higher vs Equal/Lower NRS in babies born at ≥26 weeks].

Both sets of subgroup analyses will follow a similar analytical plan as the primary analyses, including similar covariates for each outcome as long as P-value and multicollinearity conditions are met.