Abstract
Background:
Off-label treatment of extremely preterm infants with diuretics and inhaled corticosteroids (ICS) for evolving bronchopulmonary dysplasia (BPD) is common. Their effectiveness in reducing mortality or BPD severity, and optimal treatment timing, are unclear.
Objectives:
To determine if diuretic treatment or ICS administration for infants with early evolving (between 10-27 days postnatal) and progressively evolving (28 days-36 weeks postnatal) BPD are independently associated with reduced mortality and moderate or severe BPD at 36 weeks postmenstrual age (PMA).
Methods:
We examined neonates born before 28 weeks’ gestation and admitted to neonatal intensive care units on postnatal day 0 between 2006-2016 using data collected during routine care recorded within the Pediatric Health Information System (PHIS). An early evolving BPD cohort consisted of infants treated with oxygen, positive pressure, or mechanical ventilation at 10 days postnatal. The progressively evolving BPD cohort consisted of infants treated with these modalities at 28 days. In new-users, we evaluated the effect of diuretic and ICS treatment on mortality or BPD severity at 36 weeks PMA, adjusting for time dependent confounding by respiratory status using marginal structural models.
Results:
Early evolving BPD was present in 10,135 patients; progressively evolving BPD in 11,728. New diuretic exposure during early evolving BPD (adjusted risk ratio [aRR]: 0.77; 95% confidence interval [CI]: 0.65, 0.93) was associated with decreased mortality or moderate/severe BPD risk. New diuretics (aRR: 0.86; 95% CI: 0.75, 0.99) during progressively evolving BPD between 28 days-36 weeks PMA were less strongly associated with mortality or moderate/severe BPD reduction. There was no strong association for ICS in patients with early evolving (aRR: 1.40; 95% CI: 0.79, 2.51) or progressively evolving BPD (aRR: 1.16; 95% CI: 0.95, 1.49).
Conclusion:
Diuretics, but not ICS, for evolving BPD were associated with mortality and BPD risk reduction.
Keywords: Bronchopulmonary Dysplasia, diuretics, inhaled corticosteroid, neonate, preterm birth, pharmacoepidemiology
BACKGROUND
Off-label treatment of preterm infants with diuretics and inhaled corticosteroids (ICS) for both evolving and established bronchopulmonary dysplasia (BPD), the major chronic lung disease of infancy and most common severe complication of early preterm birth, is common1,2 and based on the presumption that a reduction of fluid retention and lung inflammation may be therapeutic.3-5 ICS, specifically inhaled budesonide, appear effective in reducing BPD symptoms at 36-weeks postmenstrual age (PMA) when started within the first 24-hours postnatal and continued until an infant no longer needs oxygen or respiratory support or reaches 32-weeks PMA. However in the largest trial of inhaled budesonide, which contributed most of the cumulated effectiveness data, there was a concern for potentially increased mortality in treated infants.6,7 The effectiveness of these medications in reducing mortality or improving BPD-related outcomes for infants with early evolving or progressively evolving BPD, and the time points where infants with symptomatic, ongoing respiratory symptoms might be targeted, are unclear.8-10 Given this limited evidence, we and other researchers have documented widespread practice variation in diuretic and ICS treatment between clinicians at leading US children’s hospitals.1,2
The severity of BPD is by definition graded at 36-weeks PMA, a timepoint when respiratory status more closely correlates with important long-term respiratory and neurodevelopmental outcomes.11 However, the initial presentation of the disease is much earlier.12 In 2006, experts supported by the National Institutes of Health (NIH) and US Food and Drug Administration (FDA) concluded the need for further study of diuretic and ICS treatment for early evolving BPD, beginning in the second postnatal week, and treatment of progressively evolving BPD, beginning at approximately 28 days of age.13 Since then, almost no new diuretic or ICS treatment trials have been completed to evaluate the impact of diuretic and ICS treatment during these time periods in improving BPD severity at 36-weeks PMA.8,10,14 Most existing trials on diuretic and ICS treatment of BPD were a) conducted prior to the modern neonatal medicine era (>20 years ago), b) limited by sample-size, and c) primarily powered for biomarker/physiologic outcomes instead of clinical effectiveness endpoints.
Randomized trials for preterm neonates can be difficult and resource intensive. Robust cohort data would inform trial planners by better delineating optimal treatment timing for diuretic and ICS treatment randomization. Our objectives were to determine if 1) diuretic treatment of early evolving BPD (diagnosis: 10-days postnatal), compared to no treatment, and 2) progressively evolving BPD (diagnosis: 28-days postnatal), compared to no treatment, were associated with improvement in mortality and moderate or severe BPD at 36-weeks PMA and if 3) ICS administration for treatment of early evolving BPD, and 4) progressively evolving BPD were associated with improvement in mortality and moderate or severe BPD at 36-weeks PMA.
METHODS
Inclusion Criteria
We conducted a cohort investigation using data previously collected during routine clinical care and recorded within the Pediatric Health Information System (PHIS) database. We included neonates born before 28 weeks’ gestation and admitted on postnatal day 0 to neonatal intensive care units (NICUs) contributing data to PHIS from 2006-2016. Included infants had to survive at least 10-days postnatal, the time point at which diagnosis of early evolving BPD (defined: oxygen requirement or positive pressure dependence at 10-days) was determined.
Data Source
The PHIS database contained hospital administrative and billing data from 48 tertiary, children’s hospitals affiliated with the Children’s Hospital Association. PHIS includes discrete data entry, diagnosis codes, and daily patient billing data from each participating hospital and tracks hospital data quality. In addition to demographic data, PHIS contains billing record data of medication administration and respiratory treatments for each day of an infant’s hospitalization, recorded as Clinical Transaction Classification (CTC) codes, a common system for charge-level data within PHIS.
Gestational age
Gestational age (GA) was determined from several key variables in PHIS. These included GA from the medical record and GA reported as an ICD code inclusive of a GA window spanning two-week intervals. Since GA was used to define timing of BPD outcomes, an estimated GA week was required. Approximately 17% of infants were missing GA or had information only for a GA window spanning 2 weeks. Within this window we imputed the completed GA week. Fortunately, 94% had information on the GA window or birthweight. Approximately half had both.
Designation of early evolving and progressively evolving BPD cohorts
BPD is an evolving disease that can first be detected early during hospitalization. An infant’s level of respiratory support is the driving clinical predictor of BPD severity after the first week.12 We based our early evolving and progressively evolving BPD time-intervals on the NIH and FDA-sponsored BPD expert group convened in 2006 to identify critical topics for BPD-treatment trial design.13 That group conceptualized BPD prevention and treatment in 3 stages: 1) Stage 1: perinatal and early postnatal BPD prevention up to postnatal day 7, 2) Stage 2: treatment of evolving BPD in oxygen and pressure (CPAP, ventilator) dependent infants beginning at 7-14 days postnatally, and 3) Stage 3: treatment of infants with BPD at postnatal day 28 (±7 days). We adapted those broad criteria to arbitrarily assign an early evolving BPD diagnosis at postnatal day 10 and progressively evolving BPD at day 28.
The early evolving BPD cohort consisted of all infants who were alive and treated with supplemental oxygen, any non-invasive positive pressure ventilation, or invasive mechanical ventilation (early evolving BPD) at 10 days postnatal (Figure 1). The progressively evolving BPD cohort consisted of all infants who were alive and treated with supplemental oxygen, positive pressure ventilation or mechanical ventilation (progressively evolving BPD) at 28 days postnatal.
Figure 1.
Timeline showing timepoints for assignment of early evolving BPD and for progressively evolving BPD diagnoses and for Bronchopulmonary Dysplasia (BPD) Severity Outcome Assessment at 36 weeks Postmenstrual Age (PMA).
Treatment Exposure
We created a binary diuretic variable that indicated by each day if one or more of these diuretics were administered: bumetanide (CTC code: 191131), chlorothiazide (191111), ethacrynic acid (191133), furosemide (191135), hydrochlorothiazide (191113), or spironolactone (191141). We created a binary ICS variable that indicated by each day if one or more of these ICS were administered: inhaled beclomethasone (154013.42), inhaled budesonide (154021.42), inhaled fluticasone (154055.42).
Treatment exposure models for diuretic exposure and for ICS exposure were fit to investigate our aims. Depending on specific inclusion and exclusion criteria present, each infant could be included in 1 or more of the 4 models. Among patients with early evolving BPD at 10-days postnatal we estimated the relationship of diuretic administration between days 10-27 postnatal, inclusive. A separate model compared ICS administration for early evolving BPD between days 10-27 postnatal. Among patients with progressively evolving BPD at 28-days postnatal we modeled the risk of BPD or death due to diuretic administration anytime from day 28 postnatal to 36 weeks PMA, inclusive, and separately modeled ICS administration for progressively evolving BPD treatment between day 28 postnatal-36 weeks PMA.
We made an a priori decision to only include new users15,16 not previously exposed to either study drug (diuretic or ICS) prior to the model start time (t0); day 10 postnatal for early evolving BPD models and day 28 for progressively evolving BPD. We did this because prior drug exposure complicates control for potential confounder variables (respiratory treatment modality), since those confounder covariates may plausibly be impacted by the prior treatment itself. In addition, the risk or benefit of receiving each specific drug (diuretic or ICS) might vary with time, with further drug continuation by prescribing clinician predicated upon prior response, a form of selection bias.15,17 Within all models, treatment was allowed to vary by day over time.
Outcome
Our primary composite outcome was mortality or moderate or severe BPD at 36-weeks PMA. BPD severity was determined from each infant’s billed respiratory support at 36-wks PMA.11,18 Infants on oxygen only at 36-weeks PMA were determined to have moderate BPD. Those on non-invasive or invasive positive pressure were diagnosed with severe BPD.
Statistical Analysis
We estimated via four separate marginal structural models (MSMs) the risk (adjusted risk ratio [aRR]) of mortality or moderate/severe BPD in infants for whom treatment was initiated, compared to those remaining untreated. In the case of diuretics and/or ICS treatment, an infant’s respiratory treatment status (i.e. no treatment, oxygen, CPAP, mechanical ventilation) may impact a clinician’s decision to treat. Medication treatment might affect respiratory status, which in turn influences the clinician’s decision to continue or cease treatment. Time dependent confounding occurs when exposures vary over time. In such a case, standard approaches for adjustment for confounding are biased. Based on our a priori concern for time dependent confounding by respiratory status and decision to account for it in our primary analysis, we created MSMs to allow adjustment for time-dependent confounding. MSMs allow for improved adjustment of confounding in those situations.19-22 MSMs provide average treatment effects rather than day-by-day effects. Our modeling assumptions are graphically illustrated via a Directed Acyclic Graph in Figure 2.
Figure 2.
This conceptual model, a Directed Acyclic Graph (DAG), assumes no unmeasured confounding when we control for measured covariates. As an illustration of the marginal structural model, exposure/treatment with diuretics/ICS is depicted for three time points (Diuretic/ICS0, at the time when early evolving BPD is diagnosed at postnatal day 10 or progressively evolving BPD is diagnosed at postnatal day 28, dependent on the respective model, through to PMA of 36-weeks (Diuretic/ICS1, Diuretic/ICS2 …...) during the postnatal period. C0, C1, C2 denote vectors of measured confounders, including those fixed in time (demographics and hospital characteristics) and time varying characteristics such as other treatments, surgeries, diagnoses, and respiratory modality type.
Fixed covariates included baseline demographic variables (sex, insurance, race, ethnicity, hospital indicator, hospital volume, birth year, delivery GA). Although we acknowledge the limitations of maternally reported race and ethnicity as reported in PHIS, we chose to include both in our final models given known associated outcomes disparities in preterm infants.23 Time-varying covariates for each model included binary (yes/no) treatment variables for oxygen, non-invasive positive-pressure, mechanical ventilation, or high-frequency ventilation for each patient for each day of the observation period (days 10-27 postnatal or 28 days-36 weeks PMA) and binary variables for diuretic and/or ICS treatment during each previous day.
We weighted each infant in each model by inverse probability of receiving the treatment they actually received based on measured covariates. To reduce variability and increase confidence interval precision, these weights (denominator) were stabilized (standardized) using a numerator of baseline variables only.19,24 Denominator model variables were all ‘baseline’ treatment values; fixed covariates (baseline demographics described above), treatment history the day prior for each modeled day, as well as cumulative follow-up days. The numerator model included all covariates, except those varying in time. Covariate groups were reduced to decrease unstable effect estimates for hospital size, race, ethnicity, and GA.
We modeled time-varying treatment since both diuretic and ICS therapy may have been administered for only part of the treatment window for each model ([10-27 days] or [28 days-36 weeks PMA]) and that diuretic treatment might be given for short or intermittent courses rather than continuously. Standardized weights were allowed to vary by treatment received each day.
The MSM for diuretic use or corticosteroid use was estimated through log-linear Poisson regression with robust variance estimates, weighted on each day by the stabilized weight with adjustment for the time intercept and fixed time covariates.25 The positivity assumption, that all patients had a positive probability of receiving treatment based on observed covariates was checked graphically. We assessed how close the mean and median of the estimated standardized weights were to 1.
To determine if receiving both medications (diuretics and ICS) modified the effect of either drug, we investigated the potential interaction effect between diuretics and ICS treatment. We weighted each MSM by the product of weights developed for each treatment and calculated the relative excess risk due to interaction on the additive scale (RERI).26 As a sensitivity measure, we calculated the e-value for each model to estimate how strong unmeasured confounding variable(s) would have to be, over and above the adjustment for observed confounders, to negate the observed results”.27,28
Missing data
The only key covariate with missing data was GA. GA was imputed through multiple imputation which included GA window and birthweight, along with Apgar score, sex, race, ethnicity, birth hospital, birth year, admission day, length of stay, insurance status, and outcomes. Fifty imputations were generated through a pattern mixture model, and estimates combined across imputations using Rubin’s rules.29
Ethics Approval
The study was approved by the Nationwide Children’s Hospital Institutional Review Board.
RESULTS
Treatment of Early evolving BPD from 10-27 days postnatal
Of 14,912 identified infants born before 28 weeks gestation and admitted on day 0 postnatal, early evolving BPD occurred in 13,667 patients at postnatal day 10; 10,135 were untreated with either drug. Of these infants, 4783 (47.2%) were first treated with diuretics between 10 and 27 postnatal days and 351 (3.5%) were first treated with ICS between 10 and 27 days. By 36 weeks PMA, 4494 (44.3%) had moderate or severe BPD or died (n=586 died); 4254 (43.5%) who were not treated with steroids, 240 (68.4%) who were ICS treated, 1726 (32.3%) not treated by diuretics and 2768 (57.9%) who were treated with diuretics (Table 1). Kaplan-Meier curves display cumulative survival probabilities for the 10-27 days (eFigure 1) and 28 days-36 weeks PMA (eFigure 2) subcohorts.
Table 1.
Characteristics of Early Evolving BPD (10 days postnatal) by Treatment (N=10,135)
| No Diuretic exposure (N=5352) |
Any Diuretic Exposure (N=4783) |
No ICS Exposure (N=9784) |
Any ICS Exposure (N=351) |
|
|---|---|---|---|---|
| Gestational Age a | ||||
| 22 | 17 (0.3%) | 31 (0.7%) | 46 (0.5%) | 2 (0.6%) |
| 23 | 138 (2,6%) | 330 (6.9%) | 446 (4.6%) | 22 (6.3%) |
| 24 | 373 (7.0%) | 815 (17.0%) | 1131 (11.6%) | 57 (16.2%) |
| 25 | 557 (10.4%) | 789 (16.5%) | 1276 (13.0%) | 70 (19.9%) |
| 26 | 795 (14.9%) | 816 (17.1%) | 1551 (15.9%) | 60 (17.1%) |
| 27 | 1143 (21.4%) | 695 (14.5%) | 1778 (18.2%) | 60 (17.1%) |
| 28 | 1441 (26.9%) | 584 (12.2%) | 1988 (20.3%) | 37 (10.5%) |
| Missing | 888 (16.6%) | 723 (15.1%) | 1568 (16.0%) | 43 (12.3%) |
| Admission day of life, median (p25, p75) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) |
| Birthweight (g), median (p25, p75)b | 940 (760, 1110) | 785 (641, 953) | 870 (690, 1049) | 790 (629, 965) |
| Year of Birth | ||||
| 2006-2008 | 714 (13.3%) | 843 (17.6%) | 1461 (14.9%) | 96 (27.4%) |
| 2009-2012 | 2060 (38.5%) | 1982 (41.4%) | 3876 (39.6%) | 166 (47.3%) |
| 2013-2016 | 2578 (48.2%) | 1958 (40.9%) | 4447 (45.5%) | 89 (25.4%) |
| Sex | ||||
| Female | 2649 (49.5%) | 2218 (46.4%) | 4703 (48.1%) | 164 (46.7%) |
| Race and Ethnicity | ||||
| Asian Non-Hispanic | 134 (2.5%) | 112 (2.3%) | 239 (2.4%) | 7 (2.0%) |
| Black Non-Hispanic | 1375 (25.7%) | 1172 (24.5%) | 2470 (25.3%) | 77 (21.9%) |
| Hispanic | 751 (14.0%) | 777 (16.3%) | 1477 (15.1%) | 51 (14.5%) |
| Other racec Non-Hispanic | 895 (16.7%) | 871 (18.2%) | 1688 (17.3%) | 78 (22.2%) |
| White Non-Hispanic | 2197 (41.1%) | 1851 (38.7%) | 3910 (40.0%) | 138 (39.3%) |
| Insurance/Payor | ||||
| Private | 1880 (35.1%) | 1668 (34.9%) | 3459 (35.4%) | 89 (25.4%) |
| Public | 3255 (60.8%) | 2915 (61.0%) | 5935 (60.7%) | 235 (67.0%) |
| Other | 217 (4.1%) | 200 (4.2%) | 390 (4.0%) | 27 (7.7%) |
| Hospital Volume d | ||||
| <100 | 50 (0.9%) | 40 (0.8%) | 88 (0.9%) | 2 (0.6%) |
| 100-249 | 125 (2.3%) | 122 (2.6%) | 244 (2.5%) | 3 (0.9%) |
| 250-499 | 359 (6.7%) | 448 (9.4%) | 796 (8.1%) | 11 (3.1%) |
| 500-749 | 894 (16.7%) | 887 (18.5%) | 1700 (17.4%) | 81 (23.1%) |
| ≥750 | 3924 (73.3%) | 3286 (68.7%) | 6956 (71.1%) | 254 (72.4%) |
Completed week
Birthweight missing for 506 individuals
Other race category includes American Indian or Alaskan Native, Pacific Islander, other race, or mixed race
Cohort patients per hospital
We found exposure to diuretics was associated with an average decreased risk of mortality or BPD by 36 weeks PMA (adjusted risk ratio [aRR]: 0.77 [95% CI: 0.65, 0.93]). Since some standardized weights were large, weights were truncated at various percentiles ranging at the 95th and 5th percentiles as well as values of 50 and 0.02 (Table 2).
Table 2:
Risk of BPD or Death at 36 weeks postmenstrual age among infants with early evolving or progressively evolving BPD
| MSM: stabilized weights truncated at 95th and 5th percentile [aRR, 95% CI] |
MSM: stabilized weights truncated at 50 and 0.02 [aRR, 95% CI] |
Naïve model: Any exposure and fixed baseline confounding [aRR, 95% CI] |
Naïve model: Any exposure no adjustment [aRR, 95% CI] |
|
|---|---|---|---|---|
| Early evolving BPD a | ||||
| Diuretic Exposure | 0.77 (0.65, 0.93) | 0.81 (0.68, 0.97) | 1.69 (1.61, 1.77) | 1.79 (1.71, 1.88) |
| ICS Exposure | 1.40 (0.79, 2.51) | 0.80 (0.36, 1.77) | 1.30 (1.19, 1.41) | 1.57 (1.46, 1.69) |
| Progressive evolving BPD b | ||||
| Diuretic Exposure | 0.86 (0.75, 0.99) | 0.87 (0.77, 0.98) | 2.13 (2.02, 2.25) | 2.21 (2.10, 2.33) |
| ICS Exposure | 1.16 (0.95, 1.49) | 1.25 (1.01, 1.55) | 1.53 (1.48, 1.58) | 1.64 (1.59, 1.70) |
Early evolving BPD is defined as: all infants who were alive and treated with supplemental oxygen, positive pressure ventilation or mechanical ventilation (early evolving BPD) at 10 days postnatal; adjusted for fixed baseline demographic covariates, as well as time varying treatment of ICS or diuretic, oxygen support, positive pressure, mechanical ventilation, or high frequency ventilation in MSM.
Progressively evolving BPD is defined as: all infants who were alive and treated with supplemental oxygen, positive pressure ventilation or mechanical ventilation (progressively evolving BPD) at 28 days postnatal; adjusted for fixed baseline demographic covariates, as well as time varying treatment of ICS or diuretic, oxygen support, positive pressure, mechanical ventilation, or high frequency ventilation in MSM.
The e-value27 for association between diuretic use for early evolving BPD and mortality or BPD was 1.92, implying that unmeasured confounder(s) associated with both diuretic exposure and the mortality or moderate/severe BPD outcome by an relative risk of 1.92, above currently included confounders, might alternatively explain away the effect estimate.
After adjusting for fixed and time varying confounders via MSMs, we found early ICS exposure for children with early evolving BPD at 10 days, was not associated with decreased mortality and BPD risk by 36 weeks PMA (aRR: 1.40, 95% CI: 0.79, 2.51). Due to larger stabilized weights, weights were truncated at the 95th and 5th percentiles where the mean weight was closer to 1 (Table 2). The e-value for the association between ICS use and the outcome of mortality or BPD was 2.151.
We investigated the potential interaction of effect between diuretics and ICS treatment. MSMs, weighted by the product of the weights developed for each treatment, did not provide strong evidence of interaction between treatments (RERI: −0.31, 95% CI: −1.67, 1.05).
Treatment of Progressively evolving BPD at 28 days postnatal-36 weeks PMA
Of the 11,728 surviving patients with progressively evolving BPD untreated with either drug at day 28, 7825 (66.6%) were ever treated with diuretics; 2031 (17.3%) were ever treated with ICS between day 28 postnatal-36 weeks PMA. By 36 weeks PMA, 6152 (52.5%) had moderate or severe BPD or died (n=387 died). These events comprise 4577 (47.2%) not treated with steroids, 1575 (77.6%) treated with ICS, 1131 (29.0%) not treated with diuretics and 5021 (64.2%) treated with diuretics (Table 3).
Table 3.
Characteristics of Progressively Evolving BPD (28 days postnatal) by Treatment (N=11,728)
| No Diuretic exposure (N=3903) |
Any Diuretic Exposure (N=7825) |
No ICS Exposure (N=9697) |
Any ICS Exposure (N=2031) |
|
|---|---|---|---|---|
| Gestational Age a | ||||
| 22 | 10 (0.3%) | 45 (0.6%) | 36 (0.4%) | 19 (0.9%) |
| 23 | 75 (1.9%) | 541 (6.9%) | 437 (4.5%) | 179 (8.8%) |
| 24 | 248 (6.3%) | 1486 (19.0%) | 1256 (13.0%) | 478 (23.5%) |
| 25 | 429 (11.0%) | 1388 (17.7%) | 1429 (14.7%) | 388 (19.1%) |
| 26 | 663 (17.0%) | 1303 (16.7%) | 1641 (16.9%) | 325 (16.0%) |
| 27 | 867 (22.2%) | 1109 (14.2%) | 1747 (18.0%) | 229 (11.3%) |
| 28 | 996 (25.5%) | 841 (10.8%) | 1683 (17.4%) | 154 (7.6%) |
| Missing | 618 (15.8%) | 1112 (14.2%) | 1468 (15.1%) | 259 (12.8%) |
| Admission day of life, median (p25, p75) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) |
| Birthweight (g), median (p25, p75)b | 920 (750, 1080) | 766 (630, 936) | 840 (670, 1010) | 730 (606, 882) |
| Year of Birth | ||||
| 2006-2008 | 616 (15.7%) | 1588 (20.3%) | 1794 (18.5%) | 409 (20.1%) |
| 2009-2012 | 1558 (39.9%) | 3186 (40.7%) | 3871 (39.9%) | 873 (43.0%) |
| 2013-2016 | 1730 (44.3%) | 3051 (39.0%) | 4032 (41.6%) | 749 (36.9%) |
| Sex | ||||
| Female | 1994 (51.1%) | 3549 (45.4%) | 4652 (48.0%) | 891 (43.9%) |
| Race and Ethnicity | ||||
| Asian Non-Hispanic | 109 (2.8%) | 187 (2.4%) | 256 (2.7%) | 40 (2.0%) |
| Black Non-Hispanic | 863 (22.1%) | 1946 (24.9%) | 2333 (24.1%) | 476 (23.4%) |
| Hispanic | 688 (17.6%) | 1390 (17.4%) | 1647 (17.0%) | 401 (19.7%) |
| Other racec Non-Hispanic | 644 (16.5%) | 1339 (17.1%) | 1628 (16.8%) | 355 (17.5%) |
| White Non-Hispanic | 1599 (41.0%) | 2993 (38.3%) | 3833 (39.5%) | 759 (37.4%) |
| Insurance/Payor | ||||
| Private | 1327 (34.0%) | 2675 (34.2%) | 3415 (35.2%) | 587 (28.9%) |
| Public | 2398 (61.4%) | 4767 (60.9%) | 5841 (60.2%) | 1324 (65.2%) |
| Other | 178 (4.6%) | 383 (4.9%) | 441 (4.6%) | 120 (5.9%) |
| Hospital Volume d | ||||
| <100 | 27 (0.7%) | 56 (0.7% | 73 (0.8%) | 10 (0.5%) |
| 100-249 | 133 (3.4%) | 195 (2.5%) | 303 (3.1%) | 25 (1.2%) |
| 250-499 | 263 (6.7%) | 774 (9.9%) | 844 (8.7%) | 193 (9.5%) |
| 500-749 | 718 (18.4%) | 1465 (18.7%) | 1693 (17.4%) | 490 (24.1%) |
| ≥750 | 2762 (70.8%) | 5335 (68.2%) | 6784 (70.0%) | 1313 (64.7%) |
Completed week
Birthweight missing for 724 individuals
Other race category includes American Indian or Alaskan Native, Pacific Islander, other race, or mixed race
Cohort patients per hospital
After adjusting for fixed and time varying confounders via MSMs, we found diuretic exposure which may vary day to day in this period (≥28 days postnatal-36 weeks PMA) for survivors with progressively evolving BPD at 28 days, was associated with mortality or BPD at 36 weeks PMA (aRR: 0.86, [95% CI: 0.75, 0.99]) (Table 2). Stabilized weights truncated at the 95th and 5th percentiles were highly variable. Results were similar when weights were truncated at 50 and 0.02 (aRR: 0.87, 95% CI: 0.77, 0.98), approximately the 94th and 6th percentiles of the distribution. The e-value for association between diuretic use for early evolving BPD and mortality or BPD was 1.60.
After adjusting for fixed and time varying confounders, the effect of new exposure to ICS on death or BPD at 36 weeks PMA was variable (aRR: 1.16, [95% CI: 0.95, 1.49]).
When we investigated for potential effect modification (interaction) between concurrent diuretic and ICS treatment using MSMs weighted by the product of the weights for each treatment, we found no strong evidence of synergy between treatments (RERI: 0.19, 95% CI: −0.14, 0.53). Interaction weights were truncated at a maximum of 50 to avoid extreme weights.
Naïve estimates without adjustment for time-dependent confounding or time-varying treatment
As a naïve comparison, we fit models adjusted for race, ethnicity, payor, sex, hospital size, and hospital indicator without accounting for time varying confounding and varied treatment initiation timing. These models assume continued, sustained treatment during each exposure period, when in practice it may vary. Given this assumption, we found diuretic (aRR 1.69 [1.61, 1.77]) and ICS (aRR 1.30 [1.19, 1.41]) treatments between 10-27 days for infants with early evolving BPD were associated with increased death or moderate/severe BPD risk. For infants with progressively evolving BPD at 28-days, diuretic (aRR 2.13; 95% CI: 2.02, 2.25) and ICS (aRR 1.53; 95% CI: 1.49, 1.58) between 28-days-36 weeks PMA were associated with increased death or moderate/severe BPD risks.
COMMENT
Principal findings
Treatment of infants requiring oxygen or non-invasive or invasive positive-pressure ventilation (early evolving BPD) at 10-days postnatal with diuretics was associated with reduced mortality or moderate/severe BPD risk at 36-weeks PMA. We found a similar association with diuretics for infants with progressively evolving BPD at 28-days postnatal. There was no strong association between ICS and outcomes for either cohort. We detected no evidence of interaction of effect between diuretics and ICS at either 10-27 days or 28 days-36 weeks PMA.
Strengths of the study
Strengths of our study include a large sample size of <28 weeks’ gestation neonates, availability of daily pharmacological and respiratory treatment data in PHIS, and the ability to adjust for time-varying treatment and time-varying confounding by respiratory support modality, a respiratory disease marker and the best-known clinical predictor, between days 7-27 postnatal, of BPD severity.12
Limitations of the data
Given our observational design, our study was limited by the possibility of unmeasured confounders, including potential inability to fully adjust for baseline illness. We did not have information on inborn vs outborn status, nor variables indicating maternal exposures. Given analytical complexity we did not include variables for other potential treatment exposures. Among infants with early evolving BPD, we calculated e-values of 1.92 for the association between diuretic use and our primary outcome. It is plausible that unmeasured confounder(s) associated with either medication and the mortality or moderate/severe BPD outcome, might exceed these values. The PHIS database records daily (yes/no) administration for diuretics and ICS rather than specific dose. Should an infant receive higher than the average, once per day dosing strategy most common for diuretics31 or ICS, that higher daily dose would remain uncaptured. Therefore, our findings of protective effects on mortality or moderate/severe BPD from diuretic of infants with early evolving BPD between days 10-27 postnatal should be interpreted with a degree of caution. Additionally, we did not evaluate impact of specific respiratory modality at treatment, cumulative drug exposure, or other diuretic and ICS associated outcomes including neurodevelopment, growth, or bone health. These should prompt future investigation.
Interpretation
These findings, including that nearly 50% of infants receive early diuretic treatment, are important because randomized trial data on the effectiveness of diuretics or ICS to treat evolving BPD are extremely limited, despite both medications being commonly used long-term in patients with early evolving and progressively evolving BPD with known potential adverse reactions associated with both drugs.1,2Although our results may prompt clinicians already inclined to treat with diuretics for reducing BPD severity to begin treatment between 10-27 days postnatal, the major benefit from our investigation is potential contribution to future randomized trial planning. Given the expense and extensive multicenter collaboration needed to conduct trials for <28 weeks’ gestation preterm infants, these data demonstrating the potential effectiveness of diuretic treatment at an early 10-27 days postnatal randomization time interval for reducing BPD severity is informative. Targeting symptomatic patients in earlier stages of chronic lung disease (early evolving BPD at postnatal day 10) would focus treatment on infants most likely to suffer from moderate/severe BPD while avoiding potential treatment harm for infants at lower risk. Diuretics, especially loop diuretics, are associated with electrolyte abnormalities, reduced bone density, and renal calcinosis.1 ICS, although targeted to respiratory epithelium, are still absorbed systemically. Long-term ICS exposure has been associated with growth delays in older children.30 These effects have not been fully studied in extremely preterm neonates, but given immaturity of their developing organ systems during extremely rapid growth there is reason for concern. If either drug is ultimately found effective for improving BPD severity, an optimal strategy will be to maximize benefits and minimize risks by targeting delivery to infants most likely to achieve improved outcomes.
Conclusions
When we adjusted for time-dependent confounding by each infant’s daily respiratory treatment modality via MSMs, along with other potential measured confounders, we found treatment of infants requiring oxygen or non-invasive or invasive positive-pressure ventilation (early evolving BPD) at 10-days postnatal with diuretics was associated with reduced mortality or moderate/severe BPD risk at 36-weeks PMA. We found a similar association for diuretic treatment for those infants with progressively evolving BPD at 28-days postnatal. We found neither evidence of benefit for ICS at either time interval, nor any evidence of a combined effect when infants received both medications at 10-27 days or 28 days-36 weeks PMA. The potential for residual confounding by severity of illness remains. Our results should be treated with a degree of caution. An early randomization time interval at 10-27 days postnatal in future trials of diuretics for patients with early evolving BPD, prior to the onset of other chronic neonatal comorbidities, appears promising for evaluating potential improvements in mortality and BPD severity.
Supplementary Material
eFigure 1. Survival Probability without Death or BPD Over Time for Early Evolving BPD Cohort
eFigure 2. Survival Probability without Death or BPD Over Time for Progressively Evolving BPD Cohort
SYNOPSIS.
Study Question:
Is diuretic or inhaled corticosteroid treatment (ICS) for early evolving bronchopulmonary dysplasia (BPD) at 10-27 days postnatal, and progressively evolving BPD at 28 days-36 weeks postmenstrual age (PMA), associated with improvement in mortality and BPD severity at 36 weeks PMA, and if so, which treatment timepoint appears optimal for each drug?
What’s already known:
Diuretic and ICS treatments of evolving BPD are common, but trial evidence is limited and has not demonstrated clear mortality or longer-term respiratory benefits from diuretic or ICS treatments between 10 days-36 weeks PMA.
What this study adds:
Diuretics, but not ICS, for evolving BPD were associated with reduced mortality and BPD. These data should inform trial design but not modify current clinical practice given possibility of residual confounding.
FUNDING STATEMENT
This investigation was supported by grant R03HL140272 from the National Heart, Lung, and Blood Institute (NHBLI) at the National Institutes of Health. JLS is currently funded by NHLBI grants R01HL145032 and UH3HL161338 and NIH grant UG3OD035536. EMH is currently funded by National Institute of Neurological Disorders and Stroke (NINDS) grant 5UG3NS117844.
Role of the Funder/Sponsor:
The National Heart, Lung, and Blood Institute had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
CONFLICT OF INTEREST STATEMENT
Dr. Slaughter receives supplementary grant support from Abbott for the PIVOTAL randomized trial for treatment of patent ductus arteriosus in preterm infants, which is also NHLBI-sponsored (UH3HL161338). Dr. Klebanoff and Dr. Hade report no conflicts of interest.
Footnotes
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
REFERENCES
- 1.Slaughter JL, Stenger MR, Reagan PB. Variation in the use of diuretic therapy for infants with bronchopulmonary dysplasia. Pediatrics. 2013;131(4):716–723. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Slaughter JL, Stenger MR, Reagan PB, Jadcherla SR. Utilization of inhaled corticosteroids for infants with bronchopulmonary dysplasia. PLoS One. 2014;9(9):e106838. doi: 10.1371/journal.pone.0106838 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Clouse BJ, Jadcherla SR, Slaughter JL. Systematic Review of Inhaled Bronchodilator and Corticosteroid Therapies in Infants with Bronchopulmonary Dysplasia: Implications and Future Directions. PloS one. 2016;11(2):e0148188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Jensen EA. Prevention of Bronchopulmonary Dysplasia: A Summary of Evidence-Based Strategies. Neoreviews. Apr 2019;20(4):e189–e201. doi: 10.1542/neo.20-4-e189 [DOI] [PubMed] [Google Scholar]
- 5.Thomas W, Speer C. Nonventilatory strategies for prevention and treatment of bronchopulmonary dysplasia–what is the evidence? Neonatology. 2008;94(3):150–159. [DOI] [PubMed] [Google Scholar]
- 6.Bassler D, Plavka R, Shinwell ES, et al. Early inhaled budesonide for the prevention of bronchopulmonary dysplasia. N Engl J Med. 2015;373(16):1497–1506. [DOI] [PubMed] [Google Scholar]
- 7.Shah VS, Ohlsson A, Halliday HL, Dunn M. Early administration of inhaled corticosteroids for preventing chronic lung disease in very low birth weight preterm neonates. Cochrane Database Syst Rev. 2017. Jan 4;1(1):CD001969. doi: 10.1002/14651858.CD001969.pub4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Onland W, Offringa M, van Kaam A. Late (≥ 7 days) inhalation corticosteroids to reduce bronchopulmonary dysplasia in preterm infants. Cochrane Database Syst Rev. 2022. Dec 15;12(12):CD002311. doi: 10.1002/14651858.CD002311.pub5. [DOI] [PubMed] [Google Scholar]
- 9.Stewart A, Brion LP. Intravenous or enteral loop diuretics for preterm infants with (or developing) chronic lung disease. Cochrane Database Syst Rev. 2011. Sep 7;2011(9):CD001453. doi: 10.1002/14651858.CD001453.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Stewart A, Brion LP, Ambrosio-Perez I. Diuretics acting on the distal renal tubule for preterm infants with (or developing) chronic lung disease. Cochrane Database Syst Rev. 2011. Sep 7;2011(9):CD001817. doi: 10.1002/14651858.CD001817.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ehrenkranz RA, Walsh MC, Vohr BR, et al. Validation of the National Institutes of Health consensus definition of bronchopulmonary dysplasia. Pediatrics. Dec 2005;116(6):1353–60. [DOI] [PubMed] [Google Scholar]
- 12.Laughon MM, Langer JC, Bose CL, et al. Prediction of bronchopulmonary dysplasia by postnatal age in extremely premature infants. Am. J. Respir. Crit. Care Med 2011;183(12):1715–1722. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Walsh MC, Szefler S, Davis J, et al. Summary proceedings from the bronchopulmonary dysplasia group. Pediatrics. Mar 2006;117(3 Pt 2):S52–6. doi: 10.1542/peds.2005-0620I [DOI] [PubMed] [Google Scholar]
- 14.Stewart A, Brion L. Routine use of diuretics in very-low birth-weight infants in the absence of supporting evidence. J Perinatol. 2011;31(10):633–634. [DOI] [PubMed] [Google Scholar]
- 15.Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003;158(9):915–920. [DOI] [PubMed] [Google Scholar]
- 16.Hernán MA, Alonso A, Logan R, et al. Observational studies analyzed like randomized experiments: an application to postmenopausal hormone therapy and coronary heart disease. Epidemiology. 2008;19(6):766. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. Sep 2004;15(5):615–25. doi: 10.1097/01.ede.0000135174.63482.43 [DOI] [PubMed] [Google Scholar]
- 18.Slaughter JL, Reagan PB, Newman TB, Klebanoff MA. Comparative Effectiveness of Nonsteroidal Anti-inflammatory Drug Treatment vs No Treatment for Patent Ductus Arteriosus in Preterm Infants. JAMA Pediatr. Jan 03 2017:e164354. doi: 10.1001/jamapediatrics.2016.4354 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Herńan M, Robins J. Causal Inference: What If. Hernán MA, Robins JM. Causal Inference: What If. Boca Raton: Chapman & Hall/CRC, 2020. [Google Scholar]
- 20.Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000:550–560. [DOI] [PubMed] [Google Scholar]
- 21.Bodnar LM, Davidian M, Siega-Riz AM, Tsiatis AA. Marginal structural models for analyzing causal effects of time-dependent treatments: an application in perinatal epidemiology. Am J Epidemiol. May 15 2004;159(10):926–34. doi: 10.1093/aje/kwh131 [DOI] [PubMed] [Google Scholar]
- 22.Williamson T, Ravani P. Marginal structural models in clinical research: when and how to use them? Nephrol Dial Transplant. Apr 1 2017;32(suppl_2):ii84–ii90. doi: 10.1093/ndt/gfw341 [DOI] [PubMed] [Google Scholar]
- 23.Sigurdson K, Mitchell B, Liu J, et al. Racial/Ethnic Disparities in Neonatal Intensive Care: A Systematic Review. Pediatrics. Aug 2019;144(2)doi: 10.1542/peds.2018-3114 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. Sep 15 2008;168(6):656–64. doi: 10.1093/aje/kwn164 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Richardson DB, Kinlaw AC, MacLehose RF, Cole SR. Standardized binomial models for risk or prevalence ratios and differences. Int J Epidemiol. Oct 2015;44(5):1660–72. doi: 10.1093/ije/dyv137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.TL Lash TV, Haneuse S KR. Modern Epidemiology, 4th edition. Wolters Kluwer, 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value. Ann Intern Med. Aug 15 2017;167(4):268–274. doi: 10.7326/m16-2607 [DOI] [PubMed] [Google Scholar]
- 28.Haneuse S, VanderWeele TJ, Arterburn D. Using the E-Value to Assess the Potential Effect of Unmeasured Confounding in Observational Studies. JAMA. Feb 12 2019;321(6):602–603. doi: 10.1001/jama.2018.21554 [DOI] [PubMed] [Google Scholar]
- 29.Little Roderick JA, and Rubin Donald B.. Statistical analysis with missing data. Vol. 793. John Wiley & Sons, 2019. [Google Scholar]
- 30.Kelly HW, Sternberg AL, Lescher R, et al. Effect of inhaled glucocorticoids in childhood on adult height. N Engl J Med. 2012;367(10):904–912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Bamat NA, Thompson EJ, Greenberg RG, et al. Association between postmenstrual age and furosemide dosing practices in very preterm infants. J Perinatol. Apr 2022;42(4):461–467. doi: 10.1038/s41372-022-01320-w [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eFigure 1. Survival Probability without Death or BPD Over Time for Early Evolving BPD Cohort
eFigure 2. Survival Probability without Death or BPD Over Time for Progressively Evolving BPD Cohort