Abstract
Background
The inhaled vasodilators nitric oxide and epoprostenol may be initiated to improve oxygenation in mechanically ventilated patients with severe acute respiratory failure (ARF); however, practice patterns and head-to-head comparisons of effectiveness are unclear.
Research Question
What are the practice patterns and comparative effectiveness for inhaled nitric oxide and epoprostenol in severe ARF?
Study Design and Methods
Using a large US database (Premier Healthcare Database), we identified adult patients with ARF or ARDS who were mechanically ventilated and started on inhaled nitric oxide, epoprostenol, or both. Leveraging large hospital variation in the choice of initial inhaled vasodilator, we compared the effectiveness of inhaled nitric oxide with that of epoprostenol by limiting analysis to patients admitted to hospitals that exclusively used either inhaled nitric oxide or epoprostenol. The primary outcome of successful extubation was modeled using multivariate Fine-Grey competing risk (death or hospice discharge) time-to-event models.
Results
Among 11,200 patients (303 hospitals), 6,366 patients (56.8%) received inhaled nitric oxide first, 4,720 patients (42.1%) received inhaled epoprostenol first, and 114 patients (1.0%) received both therapies on the same day. One hundred four hospitals (34.3%; 1,666 patients) exclusively used nitric oxide and 118 hospitals (38.9%; 1,812 patients) exclusively used epoprostenol. No differences were found in the likelihood of successful extubation between patients admitted to nitric oxide-only hospitals vs those admitted to epoprostenol-only hospitals (subdistribution hazard ratio, 0.97; 95% CI, 0.80-1.18). Also no differences were found in total hospital costs or death. Results were robust to multiple sensitivity analyses.
Interpretation
Large variation exists in the use of initial inhaled vasodilator for respiratory failure across US hospitals. Comparative effectiveness analyses identified no differences in outcomes based on inhaled vasodilator type.
Key Words: comparative effectiveness research, epoprostenol, nitric oxide, respiratory distress syndrome, respiratory insufficiency
Abbreviations: ARF, acute respiratory failure; IMV, invasive mechanical ventilation; iEpo, inhaled epoprostenol; iNO, inhaled nitric oxide; RRT, renal replacement therapy; SHR, subdistribution hazard ratio; SOFA, sequential organ dysfunction score
Graphical Abstract
Take-home Points.
Study Question: What are the practice patterns and comparative effectiveness for inhaled nitric oxide (iNO) and inhaled epoprostenol (iEpo) in severe acute respiratory failure?
Results: In a large multicenter retrospective study of US hospitals, wide institutional variation was found in the selection of iNO vs iEpo during severe acute respiratory failure. However, these two inhaled vasodilator therapies were associated with similar outcomes.
Interpretation: These findings suggest that iNO and iEpo likely can be used interchangeably during the management of acute respiratory failure in patients whom clinicians believe would benefit from inhaled vasodilators.
Acute respiratory failure (ARF) is a common cause of critical illness and is associated with high morbidity and mortality.1 Patients with severe hypoxemia in ARF require invasive mechanical ventilation (IMV) and may be refractory to evidence-based interventions (eg, prone positioning, high positive end-expiratory pressure) and delivery of maximum Fio2.2,3 Although not known to improve mortality,4,5 inhaled vasodilators—inhaled nitric oxide (iNO) and inhaled epoprostenol (iEpo)—may be added to improve oxygenation among patients with severe hypoxemia. However, national practices regarding inhaled vasodilator use in hypoxemic respiratory failure are unclear, and comparisons of effectiveness have been limited to small6 or mechanistic7,8 studies. In this study, we sought (1) to characterize practice patterns of iNO and iEpo across US hospitals among patients receiving IMV and (2) to conduct an observational comparative effectiveness study of iNO and iEpo by emulating a cluster randomized trial.9, 10, 11
Study Design and Methods
The protocol for this study was posted previously in an online repository.12 Modifications to the posted protocol are shown in e-Table 1.
Study Population
We used the Premier Healthcare Database,13 an enhanced, claims- and billing-based database consisting of a nonrandom 20% sample of US inpatient discharges. We included adult (≥ 18 years) patients hospitalized between 2016 and 2020 with ARF or ARDS identified using International Classification of Diseases, Tenth Revision, diagnosis codes that were designated as present on admission and who were started on IMV no earlier than the calendar day before hospital admission (to allow for IMV initiation in the ED). We limited the cohort to patients who received iEpo (pharmacy billing codes for parenteral epoprostenol and respiratory therapy billing codes for continuous inhaled treatment on the same calendar day [to exclude those who received epoprostenol intravenously]), iNO (respiratory care billing code for iNO), or both during IMV. To reduce the risk of including patients who received IV vasodilators and who received inhaled vasodilators for reasons other than refractory hypoxemia in ARF (eg, right-heart failure, pulmonary hypertension), we excluded patients who received inhaled vasodilators before the day of IMV initiation and those with documented diagnoses present on admission (eg, pulmonary hypertension), or procedures (eg, cardiac surgery) or medication use (eg, other pulmonary vasodilators) before or concurrent with inhaled vasodilator initiation. The full list of inclusion and exclusion criteria are shown in e-Table 2.
Exposure
We defined the exposure of interest as the first inhaled vasodilator (iEpo or iNO) used during IMV. The day of first exposure to inhaled vasodilators was set as study day 0.
Outcomes
Outcomes in the evaluation of practice patterns were first inhaled vasodilator (iEpo, iNO, or both), days from IMV initiation to vasodilator initiation, duration of vasodilator administration, proportion of patients who switched vasodilators, and proportion of patients per hospital who received each vasodilator type. The primary outcome in the comparative effectiveness analysis was likelihood of successful extubation within 28 days of study day 0. Successful extubation was defined as cessation of daily IMV billing codes at least one calendar day before discharge or death; the day of successful extubation was defined as the last day that IMV billing codes were used. Patients who died or were discharged to hospice on the same calendar day as extubation were coded as having the competing risk of death or discharge to hospice; patients who had IMV codes on the same calendar day as the day of discharge elsewhere (eg, another hospital, rehabilitation facility, home) were coded as not having successful extubation and were censored on the day of discharge. Secondary comparative effectiveness outcomes were (1) likelihood of initiation of renal replacement therapy (RRT), (2) hospital death or discharge to hospice, (3) extracorporeal membrane oxygenation initiation, (4) total costs (medication, equipment, and labor costs) of inhaled vasodilator therapy from the day of inhaled vasodilator initiation to the end of hospitalization, and (5) total costs (medication, equipment, and labor costs) for the hospitalization from the day of inhaled vasodilator initiation to the end of hospitalization. In the Premier Healthcare Database subset that includes patient encounters that have contributed laboratory and vital sign information to the database, we also assessed the median Pao2 to Fio2 ratio on the day after inhaled vasodilator initiation. This physiologic subset is a nonrandom sample of the larger Premier Healthcare Database that includes encounters providing granular laboratory and vital sign information to Premier, Inc., since 2016. The physiologic subset includes 27% of hospitals and 13% of patient encounters from the larger database.
Model Covariates
We identified patient- and hospital-level characteristics for inclusion as covariates in study models. Hospital-level characteristics were US census region,14 teaching status, hospital size, and the proportion of Medicaid-covered or uninsured hospital admissions. Patient-level characteristics were age, sex, race, insurance type, Elixhauser comorbidity score,15,16 previously validated measures of organ dysfunction present on admission,17,18 use of neuromuscular blockade medications on the day of inhaled vasodilator therapy initiation, and discharge year. In sensitivity analyses using the physiology subset, we also adjusted for the maximum scores for each component of the modified sequential organ dysfunction score (SOFA)18,19 on the day of inhaled vasodilator therapy initiation. We chose not to adjust for SOFA component scores in the primary analysis because they were calculable only in the physiologic subset. Additional details concerning exposure, outcomes, and additional data elements are included in e-Table 3.
Statistical Analysis
In the practice patterns analyses, we summarized outcomes using proportions, means, and medians overall and by hospital. We grouped hospitals based on their use of inhaled therapies as iEpo-only hospitals (hospitals in which patients receiving inhaled vasodilators only received iEpo), iNO-only hospitals (hospitals in which patients receiving inhaled vasodilators only received iNO), and dual-use hospitals (patients received either iEpo or iNO or both) and reported the percentage of patients who received each therapy stratified by hospital type.20 To quantify the degree to which admission hospital was a driver of first inhaled vasodilator choice, we used multivariate hierarchical logistic regression models for iNO initiation (reference: iEpo initiation) with admission hospital as a random intercept. From this model, we calculated the adjusted intraclass correlation coefficient: the percent of variation in inhaled vasodilator use explained by admission hospital after accounting for patient- and hospital-level characteristics.21
In the comparative effectiveness analysis, we emulated a cluster randomized (at the hospital level) clinical trial comparing initiation of iNO with iEpo during IMV among patients with ARF.9, 10, 11 Therefore, we limited our primary analysis cohort to patients admitted to iEpo-only or iNO-only hospitals. The risk of confounding by indication and selection bias were attenuated by a natural experiment that (1) limited the cohort to patients admitted to hospitals that likely only had a single inhaled vasodilator on formulary and (2) had active comparators in both arms.22
We used time-to-event, multivariate, Fine-Grey competing risk models accounting for clustering23 on admission hospital to determine adjusted subdistribution hazard ratios (SHRs) for successful extubation within 28 days from initiation of inhaled vasodilator therapy after accounting for the competing risk of hospital mortality or discharge to hospice. SHRs of > 1 identified increased hazards of successful extubation for iNO as compared with iEpo. Proportional hazards assumptions were checked by assessing for correlation between Schoenfeld-type residuals and time.24 We examined subgroups by discharge year (2016-2019 and 2020) to assess for effect modification based on the beginning of the COVID-19 pandemic.
For the outcome of RRT initiation, we used multivariate competing risk models akin to those for the primary outcome. For the outcomes of death or hospice discharge and for extracorporeal membrane oxygenation initiation, we used multivariate hierarchical logistic regression with hospital of admission as a random intercept. For outcomes of inhaled vasodilator therapy cost, total hospitalization costs, and median Pao2 to Fio2 ratio on the day after inhaled therapy initiation, we used multivariate hierarchical median regression models with hospital of admission as a random intercept.
We conducted several prespecified sensitivity analyses (rationale included in e-Table 4) to assess the robustness of our primary results including (1) additional model adjustment for SOFA component scores on the day of inhaled vasodilator initiation among patients admitted to iEpo-only or iNO-only hospitals in the physiologic subset, (2) restricting to patients initiated on inhaled vasodilators within 2 days of admission, (3) redefining iEpo-only and iNO-only centers based on exclusive use of one therapy at the ICU level rather than the hospital level, (4) including patients admitted to all hospital types (iEpo-only, iNO-only, and dual-use hospitals), and (5) including patients admitted to all hospital types with additional adjustment for SOFA component scores on the day of inhaled vasodilator initiation in the physiologic subset. In sensitivity analyses that included patients admitted to dual-use hospitals, we classified exposure status using the initial inhaled vasodilator and excluded patients who received both vasodilators on the same calendar day.
The α value was two-sided and set at 0.05. We did not adjust for multiple comparisons; thus, all secondary analyses should be viewed as hypothesis generating. R version 4.0.2 software (R Foundation for Statistical Computing) was used for analyses. The analysis code is available at https://github.com/nabosch/Bosch-Lab. This study was designated Not Human Subjects Research by Boston University’s Institutional Review Board (Identifier: H-41795).
Results
Practice Patterns
Among 895,520 adult patients with ARF who received IMV, 15,259 (1.7%) received inhaled vasodilators (including 2,947 of 26,581 patients [11.1%] with ARDS) and 11,200 met study inclusion criteria (Fig 1). Among included patients, 6,366 patients (56.8%) received iNO first, 4,720 patients (42.1%) received iEpo first, and 114 patients (1.0%) received both vasodilators on the same day. The median time from hospital admission to IMV initiation was 1 day (interquartile range [IQR], 1-2 days), and the median time from IMV initiation to inhaled vasodilator initiation was 1 day (IQR, 1-3 days). The median duration of iNO and iEpo was 2 days (IQRs, 0-6 and 1-5, respectively); 375 patients (3.3%) were switched from one inhaled vasodilator to the other or the second type was added in the first 6 days after initiation of the first inhaled vasodilator (e-Fig 1).
Figure 1.
Study flow diagram.
Among the 303 hospitals that administered at least one inhaled vasodilator, 81 hospitals (26.7%; 7,722 patients) administered both iNO and iEpo, 104 hospitals (34.3%; 1,666 patients) administered only iNO, and 118 hospitals (38.9%; 1,812 patients) administered only iEpo (Fig 2). Patients admitted to dual-use hospitals—compared with iEpo-only and iNO-only hospitals—were located more frequently in the South US region and were admitted to larger hospitals (Table 1). In the hierarchical logistic regression model for iNO initiation vs iEpo initiation, 94.7% of the variation in the choice of iNO over iEpo was explained by admission hospital alone after accounting for patient case-mix differences and hospital characteristics (e-Table 5). These findings are consistent with the concept that idiosyncratic variation in hospital formulary could emulate a cluster randomized clinical trial.
Figure 2.
Graph showing the percentage of patients with acute respiratory failure receiving invasive mechanical ventilation and an inhaled vasodilator who receive iEpo or iNO by hospital. The y-axis values above the x-axis intercept (blue) correspond to iNO percentages, and y-axis values below the x-axis intercept (red) correspond to iEpo percentages. In cases where patients received more than one inhaled vasodilator, individual total hospital percentages exceed 100. Most hospitals exclusively used iEpo (left side of figure) or iNO (right side of figure). iEpo = inhaled epoprostenol; iNO = inhaled nitric oxide.
Table 1.
Characteristics of Patients Initiated on a Single Inhaled Vasodilator by Hospital-Level Use
| Characteristic | iEpo-Only Hospitals (n = 1,812 Patients; n = 118 Hospitals) | iNO-Only Hospitals (n = 1,666 Patients; n = 104 Hospitals) | Dual-Use Hospitals (n = 7,722 Patients; n = 81 Hospitals) |
|---|---|---|---|
| Hospital-level characteristics | |||
| Discharge year | |||
| 2016 | 232 (12.8) | 346 (20.8) | 553 (7.2) |
| 2017 | 269 (14.8) | 260 (15.6) | 714 (9.2) |
| 2018 | 271 (15.0) | 236 (14.2) | 667 (8.6) |
| 2019 | 297 (16.4) | 276 (16.6) | 1,712 (22.2) |
| 2020 | 743 (41.0) | 548 (32.9) | 4,076 (52.8) |
| US census region | |||
| Midwest | 439 (24.2) | 214 (12.8) | 329 (4.3) |
| Northeast | 429 (23.7) | 276 (16.6) | 812 (10.5) |
| South | 527 (29.1) | 1,028 (61.7) | 5,379 (69.7) |
| West | 417 (23.0) | 148 (8.9) | 1,202 (15.6) |
| Hospital bed count | |||
| 0-299 | 721 (39.8) | 273 (16.4) | 1,278 (16.6) |
| 300-499 | 547 (30.2) | 413 (24.8) | 1,714 (22.2) |
| 500+ | 544 (30.0) | 980 (58.8) | 4,730 (72.3) |
| Teaching hospital | 952 (52.5) | 860 (51.6) | 5,583 (72.3) |
| Percentage of admissions with Medicaid or self-pay insurance | 35 (28-41) | 42 (29-51) | 45 (36-47) |
| Patient-level characteristics | |||
| Female sex | 698 (38.5) | 616 (37.0) | 3,064 (39.7) |
| Age, y | 58 (48-68) | 57 (43-66) | 58 (45-69) |
| Race | |||
| Asian | 42 (2.3) | 41 (2.5) | 173 (2.2) |
| Black | 255 (14.1) | 211 (12.7) | 1,161 (15.0) |
| White | 1,295 (71.5) | 1,093 (65.6) | 4,730 (61.3) |
| Other | 136 (7.5) | 209 (12.5) | 864 (11.2) |
| Unknown | 84 (4.6) | 112 (6.7) | 794 (10.3) |
| Insurance type | |||
| Commercial | 115 (6.3) | 120 (7.2) | 573 (7.4) |
| Medicaid | 767 (42.3) | 685 (41.1) | 3,151 (40.8) |
| Medicare | 789 (43.5) | 596 (35.8) | 3,148 (40.8) |
| Self-pay or charity | 79 (4.4) | 167 (1.0) | 476 (6.2) |
| Other | 62 (3.4) | 98 (5.9) | 374 (4.8) |
| Elixhauser comorbidity score15,16 | 5 (3-7) | 5 (3-6) | 4 (3-6) |
| Neuromuscular blockade use | 1,146 (63.2) | 1,101 (66.1) | 4,807 (62.3) |
| Dombrovskiy organ dysfunction17,18 on admission | |||
| Cardiovascular | 971 (53.6) | 957 (57.4) | 4,013 (52.0) |
| Neurologic | 254 (14.0) | 197 (11.8) | 1,152 (14.9) |
| Hematologic | 338 (18.7) | 379 (22.7) | 1,436 (18.6) |
| Hepatic | 146 (8.1) | 171 (10.3) | 703 (9.1) |
| Renal | 897 (49.5) | 812 (48.7) | 3,688 (47.8) |
| SOFA components18,19 on day of inhaled vasodilator therapy initiationa | |||
| Cardiovascular | 2 (2-4) | 2 (2-4) | 2 (2-4) |
| Respiratory | 3 (2-4) | 2 (0-4) | 3 (2-4) |
| Coagulation | 0 (0-1) | 0 (0-1) | 0 (0-0) |
| Hepatic | 0 (0-0) | 0 (0-1) | 0 (0-0) |
| Renal | 1 (0-2) | 1 (0-2) | 1 (0-2) |
| Pao2 to Fio2 ratio on the day of inhaled vasodilator initiationa,b | 106 (78-153) | 76 (56-132) | 117 (77-198) |
Data are presented as No. (%) or median (interquartile range). iEpo = inhaled epoprostenol; iNO = inhaled nitric oxide; SOFA = sequential organ failure assessment.
Only present in the Premier Healthcare Database subset with laboratory findings and vital signs used in sensitivity analyses (e-Fig 3 shows cohort sizes for these analyses).
Missing in 19.9% of patients in the database subset with laboratory and vital sign data.
Comparative Effectiveness
One thousand six hundred sixty-six patients were admitted to iEpo-only hospitals and 1,812 patients were admitted to iNO-only hospitals who were included in the primary comparative effectiveness analysis (Fig 1). These patients contributed 13,972 and 13,232 patient-days, respectively. Patient-level baseline characteristics, but not hospital-level characteristics, were similar between patients admitted to iNO-only and iEpo-only hospitals (Table 1). The median age of patients was 58 years (IQR, 45-67 years), and 64.6% of patients received neuromuscular blockade on the day of inhaled vasodilator therapy initiation. The most common acute organ dysfunctions present on admission were cardiovascular (55.4%) and renal (49.1%). The median time from IMV initiation to inhaled vasodilator therapy initiation was 1 day (IQR, 0-5 days) (e-Fig 2), and the median duration of both inhaled vasodilator therapies was 2 days (IQR, 0-4 days).
In the primary analysis cohort, 617 patients (37.0%) were extubated and 771 patients (46.3%) died or were discharged to hospice within 28 days from iNO initiation; of those initiated on iEpo, 629 patients (34.7%) were extubated and 887 patients (49.0%) died or were discharged to hospice within 28 days from iEpo initiation. No differences were found in the likelihood of successful extubation between treatment groups (iNO-only group adjusted SHR, 0.97; 95% CI, 0.80-1.18) (Table 2).15, 16, 17, 18 No differences were found between treatment groups for likelihood of RRT initiation (adjusted SHR for iNO, 0.69; 95% CI, 0.44-1.08), death or discharge to hospice (adjusted OR for iNO, 0.99; 95% CI, 0.77-1.29), extracorporeal membrane oxygenation initiation (adjusted OR for iNO, 2.05; 95% CI, 0.70-6.00), or median Pao2 to Fio2 ratio on the day after initiation of inhaled vasodilators (adjusted difference in medians between iNO-only and iEpo-only groups, –6 mm Hg; 95% CI, –21 to 9 mm Hg). The iNO-only group was associated with higher costs of inhaled vasodilator therapy (adjusted difference in medians, $3,255; 95% CI, $1,568-$4,942), but not total hospitalization costs (adjusted difference in medians, $1,676; 95% CI, –$7,729 to $11,081) (Table 3).15, 16, 17, 18
Table 2.
Comparative Effectiveness of iNO vs iEpo: Primary and Sensitivity Analyses
| Cohort | Treatment Assignment | Patient Days | Successful Extubation/Death or Hospice | Unadjusted |
Adjusted for Patient- and Hospital-Level Characteristicsa |
||
|---|---|---|---|---|---|---|---|
| SHR (95% CI) for Successful Extubationb | SHR (95% CI) for Death or Hospice Dischargeb | SHR (95% CI) for Successful Extubationb | SHR (95% CI) for Death or Hospice Dischargeb | ||||
| Admitted to iEpo-only or iNo-only hospital | iEpo (1,812) | 13,232 | 629/887 | Reference | Reference | Reference | Reference |
| iNo (1,666) | 13,972 | 617/771 | 1.06 (0.88-1.28) | 0.89 (0.74-1.07) | 0.97 (0.80-1.18)c | 1.00 (0.84-1.19) | |
| Physiologic subsetd | iEpo (284) | 2,072 | 112/119 | Reference | Reference | Reference | Reference |
| iNo (173) | 1,387 | 74/76 | 1.03 (0.71-1.49) | 1.02 (0.69-1.50) | 0.95 (0.68-1.33) | 1.11 (0.78-1.58) | |
| Inhaled vasodilators initiated within 2 d of admission | iEpo (716) | 4,984 | 321/285 | Reference | Reference | Reference | Reference |
| iNO (656) | 5,116 | 303/261 | 1.02 (0.84-1.24) | 0.95 (0.72-1.26) | 1.02 (0.80-1.29) | 1.05 (0.80-1.38) | |
| Admitted to iEpo-only or iNO-only ICU | iEpo (2,113) | 16,671 | 740/1024 | Reference | Reference | Reference | Reference |
| iNO (1,889) | 16,070 | 738/836 | 1.14 (0.96-1.36) | 0.86 (0.72-1.03) | 1.09 (0.91-1.30) | 0.94 (0.81-1.10) | |
| Admitted to any hospital | iEpo (4,720) | 41,009 | 1,731/2,262 | Reference | Reference | Reference | Reference |
| iNo (6,366) | 50,664 | 3,259/2,263 | 1.62 (1.27-2.06) | 0.67 (0.51-0.87) | 1.46 (1.21-1.75) | 0.72 (0.59-0.88) | |
| Physiologic subsetd | iEpo (515) | 4,052 | 224/211 | Reference | Reference | Reference | Reference |
| iNo (631) | 4,992 | 329/228 | 1.25 (0.96-1.63) | 0.82 (0.60-1.12) | 1.17 (0.93-1.47) | 0.89 (0.68-1.17) | |
Data are presented as No., unless otherwise indicated. iEpo = inhaled epoprostenol; iNO = inhaled nitric oxide; SHR = subdistribution hazard ratio.
Model covariates were US census region, hospital teaching status, hospital size, hospital proportion of patients who had Medicaid insurance or were self-pay, discharge year, age, sex, race, insurance type, Elixhauser comorbidity score,15,16 Dombrovskiy organ dysfunctions present on admission,17,18 and use of neuromuscular blockade medications on the day of inhaled vasodilator therapy initiation.
Within 28 d of inhaled vasodilator initiation.
Primary analysis.
Among patients included in the Premier Healthcare Database physiology subset with laboratory and vital sign data since 2016. Model included all patient- and hospital-level covariates listed above as well as each component of the modified sequential organ failure assessment score19 (cardiovascular, respiratory, hepatic, coagulation, renal) on the day of vasodilator initiation.
Table 3.
Comparative Effectiveness of iNO vs iEpo Secondary Outcomes
| Outcome | Cohort Size | Effect Estimatea (95% CI), iNO-Only vs iEpo-Only |
|---|---|---|
| Initiation of renal replacement therapy, adjusted SHR | 3,347b | 0.69 (0.44-1.08) |
| Death or discharge to hospice, aOR | 3,478 | 0.99 (0.77-1.29) |
| Extracorporeal membrane oxygenation, aOR | 3,295c | 2.05 (0.70-6.00) |
| Inhaled vasodilator costs, adjusted difference in median US dollars | 3,478 | 3,255 (1,568-4,942) |
| Total hospitalization costs, adjusted difference in median US dollars | 3,478 | 1,676 (–7,729 to 11,081) |
| Median Pao2 to Fio2 ratio on the day after inhaled vasodilator initiation, adjusted difference in median mm Hgd | 296 | –6 (–21 to 9) |
Data are presented as No., unless otherwise indicated. aOR = adjusted OR; iEpo = inhaled epoprostenol; iNO = inhaled nitric oxide; SHR = subdistribution hazard ratio.
Model covariates were US census region, hospital teaching status, hospital size, hospital proportion of patients who had Medicaid insurance or were self-pay, discharge year, age, sex, race, insurance type, Elixhauser comorbidity score,15,16 Dombrovskiy organ dysfunctions present on admission,17,18 and use of neuromuscular blockade medications on the day of inhaled vasodilator therapy initiation.
Among patients not receiving renal replacement therapy before inhaled vasodilator initiation.
Among patients not receiving extracorporeal membrane oxygenation before inhaled vasodilator initiation.
The model for median Pao2 to Fio2 ratio on the day after inhaled vasodilator therapy initiation was adjusted for median Pao2 to Fio2 ratio on the day of inhaled vasodilator therapy initiation, US census region, hospital teaching status, hospital size, hospital proportion of patients who had Medicaid insurance or were self-pay, discharge year, and hospital. Additional patient-level covariates were excluded to allow for model convergence.
Sensitivity and Subgroup Analyses
In the primary analysis cohort, subgroups limited to patients discharged before 2020 (n = 2,187; iNO-only group successful extubation adjusted SHR, 0.92; 95% CI, 0.75-1.13) and during 2020 (n = 1,291 [71.1% had a diagnosis of COVID-19]; iNO-only group successful extubation adjusted SHR, 1.10; 95% CI, 0.81-1.49) were similar to those in the primary analysis. Results of sensitivity analyses (Table 2, e-Fig 3) generally were similar to the primary analysis. In the sensitivity analysis including patients admitted to any hospital type (including dual-use hospitals [n = 11,086]), patients receiving iNO had higher likelihood of successful extubation (SHR, 1.46; 95% CI, 1.21-1.75) when adjusted for admission organ dysfunction, which was attenuated in the physiologic subset cohort (n = 1,146) in which electronic health data allowed for additional model adjustment for the SOFA score on the day of inhaled vasodilator initiation (SHR, 1.17; 95% CI, 0.93-1.47).
Discussion
We used a multicenter US inpatient database to examine practice patterns and outcomes for iNO and iEpo in the management of ARF. We found large hospital variation in the choice of iNO or iEpo among mechanically ventilated patients with ARF: more than two-thirds of hospitals used a single type of inhaled vasodilator, and admission hospital explained 95% of the variation in initial inhaled vasodilator type. In an emulation of a cluster-randomized trial, our comparative effectiveness study found no difference in the likelihood of successful extubation between patients admitted to hospitals that exclusively used iNO vs patients admitted to hospitals that exclusively used iEpo. These findings suggest that iEpo and iNO likely can be used interchangeably during the management of ARF in patients who clinicians think would benefit from inhaled vasodilators.
Our results should be considered in the context of existing studies. First, it is important to note that neither iNO nor iEpo have been shown to provide mortality benefit.4,5 We found that just less than 2% of patients with ARF (including 11% of those with ARDS) are administered inhaled vasodilators in contemporary practice (translating to approximately 20,000 patients per year across the United States).25 We identified large variation in the use of iNO and iEpo, with two-thirds of hospitals using only a single type of inhaled vasodilator. This large interhospital variation suggests that choice of inhaled vasodilator may be based on hospital formulary, rather than specific patient characteristics, findings consistent with results from single-center studies.26,27 Prior single-center observational studies6,28 and pilot clinical trials7,8,29 found no differences in clinical outcomes between iNO and iEpo; both therapies transiently increased oxygenation, but had no other observable beneficial effects. Our study provides stronger and more robust evidence that no differences exist in patient outcomes based on inhaled vasodilator type. Future studies should examine processes of care (eg, complexity of starting therapy) that differ between iNO and iEpo that may identify clinical scenarios where one treatment may be preferential. Compared with placebo, prior studies suggested that iNO may increase the risk of renal failure.4 In our study, iNO was not associated with increased risk of new RRT compared with iEpo.
Limited evidence28,30 previously suggested that iNO was associated with increased costs (using estimated iNO contract costs, rather than actual costs; iNO purchase costs are not publicly available), a finding that prompted some hospitals to switch to iEpo.26 Using hospital billing data to identify actual costs, we found increased inhaled vasodilator-specific costs for iNO, but no difference in total hospitalization costs. Our findings are consistent with those by Angus et al,31 who found no difference in total hospitalization costs between treatment arms of a clinical trial comparing iNO with placebo. We speculate that signals for small differences in the cost of inhaled vasodilators may be hidden by the large cost of total hospitalization in patients with severe ARF.31
Our study has several strengths. We included more than 25 times the number of patients previously studied, and thus were able to characterize use variation and to obtain more precise estimates of effectiveness than previously possible. In addition, by leveraging hospital use of a single inhaled vasodilator and by using an active treatment comparator, we were able to reduce the risk of patient-level confounding.
Our study also has limitations. The Premier Healthcare Database does not contain medication dose information that might enable better characterization of exposure. In addition, the Premier Healthcare Database includes a nonrandom sample of US hospitalizations, and the subset of patients with laboratory and vital sign data also is a nonrandom sample of the larger database; the impact of this on the external generalizability of our findings is unknown. We were unable to ascribe the specific indication for inhaled vasodilator therapy (eg, refractory hypoxemia, right ventricular failure). However, we excluded patients with diagnoses, procedures, and medications that may suggest alternative indications for inhaled vasodilators, and a high percentage of patients received concomitant neuromuscular blockade, thus increasing confidence that the study cohort included patients who received inhaled vasodilators for refractory hypoxemia in ARF. Although we were able to adjust for concurrent use of neuromuscular blockade, we were unable to adjust for other adjunct therapies, such as prone positioning or adherence to lung-protective ventilation. We did not include an exposure group with no inhaled therapy because we were concerned of the risk for confounding by indication. Therefore, our study results should not be used to make clinical decisions regarding the benefit of initiation of inhaled vasodilators.
Interpretation
In a large multicenter retrospective study, we found large variation in the use of iNO and iEpo during ARF across the United States. Leveraging this variation to emulate a cluster randomized clinical trial, we found no difference in the comparative effectiveness of iNO vs iEpo for successful extubation, death or hospital discharge, or total hospitalization costs. These results suggest that, among patients with ARF or ARDS treated with inhaled vasodilators, such as those in this cohort, iEpo and iNO can be used interchangeably.
Acknowledgments
Authors contributions: N. A. B. takes responsibility for the integrity of the work as a whole, from inception to published article. All authors substantially contributed to the conception and design of this study. N. A. B. acquired the data. All authors were involved in the interpretation of data. N. A. B. drafted the manuscript and all authors revised it critically for important intellectual content. All authors read and approved the final manuscript.
Funding/support: This study was supported by National Center for Advancing Translational Sciences (NCATS), National Institutes of Health (NIH) [Grant 1KL2TR001411]. N. A. B. is supported by the National Heart, Lung, and Blood Institute (NHLBI) and NCATS of the NIH and the US Department of Defense (DOD). A. C. L. is supported by the NHLBI, NIH. K. R. G. is supported by a Parker B. Francis Fellowship Award. H. B. G. received funding from the University of Miami Hospital and Clinics Data Analytic Research Team (UHealth-DART). H. W. is supported by the US DOD and a Canada Research Chair (CIHR). A. J. W. is supported by the NHLBI of the NIH and the US DOD. E. A. V. is supported by the Agency for Healthcare Research and Quality.
Financial/nonfinancial disclosures: None declared.
Role of sponsors: This study’s contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
Additional information: The e-Figures and e-Tables are available online under “Supplementary Data.”
Footnotes
Drs Bosch and Law contributed equally to this manuscript.
Supplementary Data
References
- 1.Vincent J.-L., Akça S., De Mendonça A., et al. The epidemiology of acute respiratory failure in critically ill patients(∗) Chest. 2002;121(5):1602–1609. doi: 10.1378/chest.121.5.1602. [DOI] [PubMed] [Google Scholar]
- 2.Guérin C., Reignier J., Richard J.-C., et al. Prone positioning in severe acute respiratory distress syndrome. N Engl J Med. 2013;368(23):2159–2168. doi: 10.1056/NEJMoa1214103. [DOI] [PubMed] [Google Scholar]
- 3.Acute Respiratory Distress Syndrome Network. Brower R.G., Matthay M.A., et al. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med. 2000;342(18):1301–1308. doi: 10.1056/NEJM200005043421801. [DOI] [PubMed] [Google Scholar]
- 4.Gebistorf F., Karam O., Wetterslev J., Afshari A. Inhaled nitric oxide for acute respiratory distress syndrome (ARDS) in children and adults. Cochrane Database Syst Rev. 2016;6:CD002787. doi: 10.1002/14651858.CD002787.pub3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Afshari A., Brok J., Møller A.M., Wetterslev J. Aerosolized prostacyclin for acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) Cochrane Database Syst Rev. 2010;8:CD007733. doi: 10.1002/14651858.CD007733.pub2. [DOI] [PubMed] [Google Scholar]
- 6.Buckley M.S., Agarwal S.K., Garcia-Orr R., Saggar R., MacLaren R. Comparison of fixed-dose inhaled epoprostenol and inhaled nitric oxide for acute respiratory distress syndrome in critically ill adults. J Intensive Care Med. 2021;36(4):466–476. doi: 10.1177/0885066620906800. [DOI] [PubMed] [Google Scholar]
- 7.Zwissler B., Kemming G., Habler O., et al. Inhaled prostacyclin (PGI2) versus inhaled nitric oxide in adult respiratory distress syndrome. Am J Respir Crit Care Med. 1996;154(6 pt 1):1671–1677. doi: 10.1164/ajrccm.154.6.8970353. [DOI] [PubMed] [Google Scholar]
- 8.Walmrath D., Schneider T., Schermuly R., Olschewski H., Grimminger F., Seeger W. Direct comparison of inhaled nitric oxide and aerosolized prostacyclin in acute respiratory distress syndrome. Am J Respir Crit Care Med. 1996;153(3):991–996. doi: 10.1164/ajrccm.153.3.8630585. [DOI] [PubMed] [Google Scholar]
- 9.Page L.C., Lenard M.A., Keele L. The design of clustered observational studies in education. AERA Open. 2020;6(3) [Google Scholar]
- 10.Hansen B.B., Rosenbaum P.R., Small D.S. Clustered treatment assignments and sensitivity to unmeasured biases in observational studies. J Am Stat Assoc. 2014;109(505):133–144. [Google Scholar]
- 11.Hernán M.A., Robins J.M. Using big data to emulate a target trial when a randomized trial is not available. Am J Epidemiol. 2016;183(8):758–764. doi: 10.1093/aje/kwv254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bosch N. Center for Open Science website; March 7, 2022. Comparative effectiveness of inhaled epoprostenol and nitric oxide in acute hypoxemic respiratory failure: a target trial emulation.https://osf.io/xbmjk/ [Google Scholar]
- 13.Premier. Premier Healthcare Database white paper: data that informs and performs Premier website. https://learn.premierinc.com/white-papers/premier-healthcaredatabase-whitepaper
- 14.United States Census Bureau 2010 Census regions and divisions of the United States. United States Census Bureau website. 2010. https://www.census.gov/geographies/reference-maps/2010/geo/2010-census-regions-and-divisions-of-the-united-states.html
- 15.Elixhauser A., Friedman B., Stranges E. Healthcare Cost and Utilization Project (HCUP) Statistical Briefs [Internet]. Agency for Healthcare Research and Quality (US); 2006. National Library of Medicine website; 2009. Statistical brief #122: septicemia in U.S. hospitals, 2009.http://www.ncbi.nlm.nih.gov/books/NBK65391/ [Google Scholar]
- 16.Quan H., Sundararajan V., Halfon P., et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–1139. doi: 10.1097/01.mlr.0000182534.19832.83. [DOI] [PubMed] [Google Scholar]
- 17.Dombrovskiy V.Y., Martin A.A., Sunderram J., Paz H.L. Rapid increase in hospitalization and mortality rates for severe sepsis in the United States: a trend analysis from 1993 to 2003. Crit Care Med. 2007;35(5):1244–1250. doi: 10.1097/01.CCM.0000261890.41311.E9. [DOI] [PubMed] [Google Scholar]
- 18.Bosch N.A., Law A.C., Rucci J.M., Peterson D., Walkey A.J. Predictive validity of the sequential organ failure assessment score versus claims-based scores among critically ill patients. Ann Am Thorac Soc. 2022;19(6):1072–1076. doi: 10.1513/AnnalsATS.202111-1251RL. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Nates J.L., Cárdenas-Turanzas M., Wakefield C., et al. Automating and simplifying the SOFA score in critically ill patients with cancer. Health Informatics J. 2010;16(1):35–47. doi: 10.1177/1460458209353558. [DOI] [PubMed] [Google Scholar]
- 20.Centers for Medicare and Medicaid Services Measure methodology. Centers for Medicare and Medicaid Services website. https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Measure-Methodology
- 21.Merlo J., Chaix B., Yang M., Lynch J., Råstam L. A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. J Epidemiol Commun Health. 2005;59(6):443–449. doi: 10.1136/jech.2004.023473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gershon A.S., Jafarzadeh S.R., Wilson K.C., Walkey A.J. Clinical knowledge from observational studies. Everything you wanted to know but were afraid to ask. Am J Respir Crit Care Med. 2018;198(7):859–867. doi: 10.1164/rccm.201801-0118PP. [DOI] [PubMed] [Google Scholar]
- 23.Do Ha I., Christian N.J., Jeong J.-H., Park J., Lee Y. Analysis of clustered competing risks data using subdistribution hazard models with multivariate frailties. Stat Methods Med Res. 2016;25(6):2488–2505. doi: 10.1177/0962280214526193. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Scrucca L., Santucci A., Aversa F. Regression modeling of competing risk using R: an in depth guide for clinicians. Bone Marrow Transplant. 2010;45(9):1388–1395. doi: 10.1038/bmt.2009.359. [DOI] [PubMed] [Google Scholar]
- 25.Kempker J.A., Abril M.K., Chen Y., Kramer M.R., Waller L.A., Martin G.S. The epidemiology of respiratory failure in the United States 2002-2017: a serial cross-sectional study. Crit Care Explor. 2020;2(6) doi: 10.1097/CCE.0000000000000128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Davis S.L., Crow J.R., Fan J.R., et al. Use and costs of inhaled nitric oxide and inhaled epoprostenol in adult critically ill patients: A quality improvement project. Am J Health Syst Pharm. 2019;76(18):1413–1419. doi: 10.1093/ajhp/zxz151. [DOI] [PubMed] [Google Scholar]
- 27.Collier T.E., Kataria V., Modrykamien A. Inhaled epoprostenol utilization pattern after implementation of an administration policy. Proc (Bayl Univ Med Cent) 2019;33(1):10–14. doi: 10.1080/08998280.2019.1668668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Torbic H., Szumita P.M., Anger K.E., Nuccio P., LaGambina S., Weinhouse G. Inhaled epoprostenol vs inhaled nitric oxide for refractory hypoxemia in critically ill patients. J Crit Care. 2013;28(5):844–848. doi: 10.1016/j.jcrc.2013.03.006. [DOI] [PubMed] [Google Scholar]
- 29.Van Heerden P.V., Blythe D., Webb S.A. Inhaled aerosolized prostacyclin and nitric oxide as selective pulmonary vasodilators in ARDS—a pilot study. Anaesth Intensive Care. 1996;24(5):564–568. doi: 10.1177/0310057X9602400510. [DOI] [PubMed] [Google Scholar]
- 30.McGinn K., Reichert M. A comparison of inhaled nitric oxide versus inhaled epoprostenol for acute pulmonary hypertension following cardiac surgery. Ann Pharmacother. 2016;50(1):22–26. doi: 10.1177/1060028015608865. [DOI] [PubMed] [Google Scholar]
- 31.Angus D.C., Clermont G., Linde-Zwirble W.T., et al. Healthcare costs and long-term outcomes after acute respiratory distress syndrome: a phase III trial of inhaled nitric oxide∗. Crit Care Med. 2006;34(12):2883–2890. doi: 10.1097/01.CCM.0000248727.29055.25. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.