The figure depicts the decrease in vitamin A use during the study period. Concurrent with this decrease, there was no change in the prevalence of death or chronic lung disease (CLD).
Tolia VN, Murthy K, McKinley PS, Bennett MM, Clark RH. The Effect of the National Shortage of Vitamin A on Death or Chronic Lung Disease in Extremely Low-Birth-Weight Infants. JAMA Pediatr. 2014;168(11):1039-1044. doi:10.1001/jamapediatrics.2014.1353
Prophylactic vitamin A supplementation has been shown to reduce the incidence of chronic lung disease or death in extremely low-birth-weight infants. Beginning in 2010, a national shortage reduced the supply of vitamin A available.
To estimate the association between vitamin A supplementation and death or chronic lung disease in the context of the recent drug shortage. Intercenter variability in vitamin A use was assessed secondarily.
Design, Setting, and Participants
Retrospective cohort study of 7925 infants with birth weights between 401 and 1000 g who were cared for in US neonatal intensive care units managed by the Pediatrix Medical Group. Infants were discharged between January 1, 2010, and June 30, 2012, and data were collected from the Pediatrix Clinical Data Warehouse. Infants who had major congenital anomalies, died during the first 3 days of life, or had missing data were excluded from the analysis.
Vitamin A supplementation.
Main Outcomes and Measures
The primary outcome was either death before hospital discharge or chronic lung disease, defined as receiving any respiratory support at 36 weeks’ corrected gestational age.
Of the 6210 eligible infants, 3011 (48.5%) experienced the primary outcome. Those who received vitamin A were more immature and more likely to receive mechanical ventilation during the first 3 days of life. During the study period, vitamin A supplementation significantly decreased (27.2% to 2.1%); however, the primary outcome was similar (48.4% to 49.5%; P = .40). Vitamin A was unrelated to death or chronic lung disease in unadjusted or multivariable analyses (relative risk [RR], 0.97; 95% CI, 0.91-1.03; P = .32) when demographic and clinical information were considered. After classifying centers by vitamin A use, the center of birth was significantly associated with the outcome, with birth in low- and medium-use centers related to a reduced likelihood of death or chronic lung disease.
Conclusions and Relevance
The occurrence of death or chronic lung disease appears unaffected by the recent shortage of vitamin A. However, the center of birth appears to be an important risk factor for these infants’ outcomes.
In extremely low-birth-weight (ELBW) infants, chronic lung disease (CLD) has been associated with cardiopulmonary morbidity, adverse neurodevelopmental outcome, and mortality.1 Multiple strategies have been investigated to reduce the risk of death or CLD.2- 6 In particular, Tyson and colleagues2 and Ambalavanan and colleagues7 demonstrated that prophylactic vitamin A supplementation modestly and safely reduced the rates of death or CLD in ELBW infants with no adverse effects at the 18- to 22-month follow-up. The use of vitamin A has increased in the past decade,8 although vitamin A use has remained variable because of the perceived unfavorable risk to benefit ratio, among other reasons detailed by Kaplan et al.9 The national supply of the injectable form of vitamin A became scarce in late 2010 and continues to be commercially unavailable.10 Correspondingly, the anticipated reduction in vitamin A use and the related effect on the prevalence of death or CLD remains uncertain.
We hypothesized that the recent national shortage in vitamin A might be associated with an increase in death or CLD before discharge from the neonatal intensive care unit (NICU). Thus, the aims of this study were to (1) examine the use of vitamin A in the participating NICUs and (2) quantify the association between vitamin A and the composite outcome of death or CLD in ELBW infants before and during the recent national shortage.
The institutional review board of the Baylor Research Institute and the Research Advisory Committee of Mednax, Inc, provided oversight for this analysis. Data were obtained from the Pediatrix Clinical Data Warehouse, which maintains deidentified clinical information, including the details of admission, clinical course, and discharge of infants cared for in more than 200 NICUs in 33 states. A prior study11 estimated that these data capture 20% of infants admitted to NICUs in the United States.
This study was a retrospective cohort analysis of inborn infants who weighed between 401 and 1000 g at birth, were cared for at a single site during their entire hospitalization, and were discharged between January 1, 2010, and June 30, 2012. Infants who had major congenital anomalies or who died within 3 days of life were excluded. These criteria were chosen to define a study population similar to that in the published clinical trial.2
The primary outcome was death or the development of CLD before NICU discharge. Chronic lung disease was defined as requiring supplemental oxygen at 36 weeks’ corrected gestation.12 Infants discharged before 36 weeks were considered to have CLD only if they were discharged while still receiving supplemental oxygen. The reported physiologic test for bronchopulmonary dysplasia was not systematically performed and thus was unavailable for this analysis.13 Secondary outcomes included death and CLD individually as well as the following prematurity-associated morbidities: any report of necrotizing enterocolitis, intraventricular hemorrhage, retinopathy of prematurity stage III or worse, and length of stay. We included suspected, medical, and surgical necrotizing enterocolitis in the term “any necrotizing enterocolitis.” Time was reported as the year and quarter of discharge for each infant.
Stratified by vitamin A supplementation, the demographic and clinical characteristics were described for the eligible infants. Maternal characteristics (eg, age, parity, race, or ethnicity), antenatal therapies (eg, prenatal care, antenatal corticosteroids), and birth characteristics (eg, gestational age at birth, small for gestational age [calculated using normative data]14) were reported. To approximate the severity of illness, additional variables were captured (eg, 5-minute Apgar score, receiving mechanical ventilation in the first 3 days of life, and surfactant administration).
Vitamin A supplementation was ascertained from the medication tables in the Pediatrix Clinical Data Warehouse. These tables are populated during the clinical encounter and contain data about the medication and number of days of therapy. However, information related to the specific dosage and methods of administration were removed during data extraction to maintain anonymity. We defined the preshortage period as the aggregate data from infants who were discharged from the start of the first quarter through the end of the fourth quarter of 2010. The shortage period was defined as data from infants discharged from the start of the fourth quarter of 2011 through the end of the second quarter of 2012.
Although each center’s identity was masked, data on the altitude, number of eligible infants, and preshortage use of vitamin A in each center were collected. We defined vitamin A use in each center as the proportion of eligible infants who were given vitamin A in the preshortage period. These centers were then grouped into quintiles of vitamin A use (none, >0-25%, >25%-50%, >50%-75%, and >75%).
Bivariate analyses were conducted to compare the demographic, antenatal, birth, and acuity measures between infants who received vitamin A and those who did not. t Tests and χ2 tests were used, as appropriate, and Wilcoxon rank sum tests were applied for variables whose distribution was nonparametric. Using the results of the bivariate analyses, variables that were significantly associated with vitamin A supplementation or were hypothesized to influence the risk of death or CLD were retained for the regression analysis. With the inclusion of vitamin A supplementation, multivariable Poisson regression models with robust error variances were constructed to quantify the association between vitamin A and the primary outcome. These models treated the center as a fixed effect and the results are reported as relative risk (RR). We also repeated this analysis using multivariable logistic regression models, in which the center was treated as a random effect to adjust for variances in the outcome both between and within centers.
We also performed an expected to observed analysis. Using a generalized linear model with a logistic link function, we fit a model to determine the factors that influence death or CLD using only data in the preshortage period. The estimated coefficients from this model were then used to estimate the probability of death or CLD for all patients during the shortage period. The sum of these probabilities is the expected number of deaths or CLD cases. A χ2 distribution was then used to evaluate the expected to observed ratio for death or CLD during the shortage period.
Analyses, conducted using SAS, version 9.3 (SAS Institute Inc), were performed by the Department of Quantitative Sciences at Baylor Health Care System.
There were 7925 ELBW infants identified; infants who had major congenital anomalies (n = 1125), died in the first 72 hours of life (n = 564), had missing CLD data at 36 weeks’ corrected gestation (n = 21), or had weight discrepancies at birth and admission (n = 5) were omitted from the analysis. Thus, 6210 infants were eligible for the study.
Of these eligible infants, 1085 (17%) received vitamin A and 3011 (48.5%) experienced the primary outcome of death or CLD. As shown in the Figure, vitamin A supplementation declined precipitously in the beginning of 2011 in the entire cohort. During the same period, the prevalence of death or CLD did not change significantly (before shortage, 48.4%; during shortage, 49.5%; P = .40).
Significant differences in the characteristics of the population were demonstrated after stratifying the eligible infants by vitamin A supplementation (Table 1). Infants who received vitamin A were slightly less mature, with lower mean birth weights, and were less frequently small for gestational age at birth. They were also more likely to be white and to have received surfactant and mechanical ventilation in the first 3 days of life. Despite these differences, the overall frequency of the primary outcome was similar regardless of vitamin A supplementation in bivariate analysis (vitamin A, 51%; no vitamin A, 48%; RR, 1.06; 95% CI, 0.99-1.13, P = .10). There were no significant differences in outcomes when CLD and death were separated or further categorized by degree of respiratory support at hospital discharge (Table 2).
In multivariable analysis, vitamin A supplementation was also not related to death or CLD after gestational age, being small for gestational age, sex, race, severity of illness, and center characteristics were considered in the regression equations (RR, 0.97; 95% CI, 0.91-1.03; P = .32).
For the expected to observed analysis, we developed a prediction model for the outcome based on the infants born in the preshortage period only (when vitamin A was available) using the following maternal and clinical risk factors: gestational age, small for gestational age status, sex, race, antenatal corticosteroid use, 5-minute Apgar score of less than 5, surfactant use, receipt of mechanical ventilation in the first 3 days of life, site altitude, and site volume. This prediction model had an area under the curve of 0.786 for predicting CLD or death in the infants born in the preshortage period. This model was then applied to each infant during the shortage period to develop an expected rate of CLD or death. The expected rate of the outcome in the shortage period was 48.4% (area under the curve, 0.780) compared with an observed rate of 48.5% with an expected to observed ratio of 1.0 (95% CI, 0.95-1.05; P = .75).
Centers were classified into quintiles according to each center’s vitamin A use before the shortage to adjust for intracenter selection bias of vitamin A treatment and to consider intercenter variation in both practices and outcomes. First, we performed a separate analysis on the subcohort of infants from centers with more than 75% preshortage vitamin A use to restrict our analysis to centers that used prophylactic vitamin A supplementation (n = 1313). Despite a steep decline in vitamin A supplementation in the centers with more than 75% vitamin A use before the shortage (from 95.4% before the shortage to 7.4% during the shortage period), the prevalence of the primary outcome in this subcohort of NICUs was similar for the entire study period (47.0% in both the preshortage and shortage periods) (Table 3). After repeating the multivariable logistic and Poisson regression analyses in this subcohort of infants born in the centers with more than 75% vitamin A use, there was a slight trend, but no significant association, between vitamin A and the primary outcomes (RR, 0.90; 95% CI, 0.81-1.00; P = .054). Of note, one center had sustained vitamin A use throughout the study period, with 100% of their study-eligible infants receiving vitamin A and accounting for most of the vitamin A use during the shortage period. This center used a vitamin A formulation provided by a compounding pharmacy instead of the commercial product during the shortage. In regression analysis of this subcohort without this center, there was still no association with the primary outcome (RR, 0.95; 95% CI, 0.86-1.06; P = .37).
As a function of vitamin A use at each center, significant intercenter variation was observed in death or CLD (P = .002). However, these differences were unrelated to the onset of the national shortage (Table 3). In reference to NICUs without vitamin A use, infants born at centers with low and modest use of vitamin A experienced a lower likelihood of the primary outcomes after adjustment for demographic, antenatal, and birth characteristics and vitamin A supplementation (none = 1 [reference]; vitamin A use >0%-25% RR, 0.90; 95% CI, 0.82-0.98; P = .02; vitamin A use >25%-50% RR, 0.88; 95% CI, 0.78-0.98; P = .02) (Table 4). Analysis using multivariable logistic regression models, in which the center was treated as a random effect, produced similar results in all cases.
Using a national, large, modern cohort of infants, this study describes the decreasing use of vitamin A in ELBW infants concurrent with a known national shortage. Preshortage vitamin A use was similar to that in previous articles,8,9 and this study contributes new information about the intercenter variability in the use of vitamin A. In this cohort, the reduction in vitamin A use was not related to a hypothesized increase in the risk of death or CLD despite prior randomized trials and meta-analyses reporting a protective effect of vitamin A.2,15 Both the regression and expected to observed analysis suggest that vitamin A did not have an effect on the risk of death or CLD.
Several factors may contribute to these findings. Recent changes in clinical management (eg, increase in noninvasive respiratory support) have been associated with the observed lower overall incidence of death or CLD.16 This cohort had high rates of exposure to antenatal corticosteroids and caffeine as well as lower incidence of CLD or death compared with the findings by Tyson et al.2 Both these factors may have reduced the clinical benefit of vitamin A and led to a smaller treatment “signal” in our data. Treatment bias, in which only sicker infants received vitamin A, is another possibility. However, we were not able to detect an effect of vitamin A in our subcohort of infants in the centers where almost all infants received vitamin A before the shortage. Finally, given the retrospective nature of the data, it is possible that another clinical intervention was introduced (such as change in oxygen parameters) during the shortage that replaced the benefit of vitamin A. We were unable to control for this type of intervention but would have expected to see some effect of this intervention in the centers with 0% vitamin A use.
These results also support prior studies17,18 that have shown that the individual center of treatment remains a significant influence on prematurity-associated outcomes, even though the precise mechanism of this effect remains uncertain. Previous studies19- 21 have shown that wide intercenter variability exists in both the incidence of CLD as well as the prevalence of being discharged while still receiving supplemental oxygen. Significant variability was observed in the primary outcome when centers were grouped by vitamin A use. In this cohort, birth in NICUs that had low or intermediate use of vitamin A before the shortage reduced the risk of death or CLD, independent of clinical predictors of the primary outcome. This study is notable for quantifying the degree of risk reduction observed when the center of birth was included in the analysis. Because vitamin A supplementation was not associated with the outcome in this cohort, stratification of sites by vitamin A use likely represents an independent proxy related to center-level practice parameters. Correspondingly, the practices and characteristics of these centers warrant further investigation to isolate the interventions that may be protective and to improve future outcomes for at-risk ELBW infants.
This study took advantage of the so-called natural experiment phenomenon and captured the actual reduction in vitamin A use concurrent with the shortage of this medication. Although the reported effect of vitamin A is modest, this study would be challenging to conduct as a prospectively randomized trial, given the barriers to restricting and randomizing infants to preclude the receipt of vitamin A. This analysis also includes a large and representative number of eligible ELBW infants (N = 6210), of whom 1085 received vitamin A. In the subcohort of infants born in centers with more than 75% vitamin A use, there were 1313 infants. This is compared with the original randomized trial, which included 807 infants.2 In addition, collected data for this cohort was analyzed during a short period of time (to minimize changes in treatment patterns) and across a heterogeneous collection of NICUs that were represented in the data set. Finally, relatively few infants were excluded for missing information.
This analysis has limitations. There may be important clinical characteristics unaccounted for in our model or further clinical changes that physicians used to compensate for the loss of vitamin A, resulting in no net change in the composite outcome. Also, the level of detail in the data set did not allow for a physiologic definition of CLD, which has been shown to reduce variability in reporting this outcome.22 However, the definition of CLD was consistent with the original clinical trials of vitamin A.2 As with similar studies using administrative databases,23 these findings may be subject to underreporting and data entry error despite internal data consistency checks and low rates of missing data. Ultimately, owing to the nature of retrospective reviews, this study is only able to generate hypotheses because of the lack of an observed effect of vitamin A in this cohort.
In this retrospective analysis of a large national neonatal database, we report a large decrease in vitamin A use that was not associated with an increase in death or CLD in ELBW infants. In examining vitamin A use by center, we identified NICUs that observed a reduced prevalence of the composite outcome in these high-risk ELBW infants. Future study will be completed to help clarify center-specific practices to guide future clinical trials or quality-improvement initiatives.
Accepted for Publication: June 11, 2014.
Corresponding Author: Veeral N. Tolia, MD, Division of Neonatology, Department of Pediatrics, Baylor University Medical Center, 3500 Gaston Ave, 3 Hoblitzelle, Dallas, TX 75246 (email@example.com).
Published Online: September 15, 2014. doi:10.1001/jamapediatrics.2014.1353.
Author Contributions: Dr Tolia had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Tolia, McKinley, Clark.
Acquisition, analysis, or interpretation of data: Tolia, Murthy, Bennett, Clark.
Drafting of the manuscript: Tolia, Murthy, Clark.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Tolia, Bennett.
Obtained funding: Tolia, McKinley.
Administrative, technical, or material support: Tolia.
Study supervision: Tolia, Murthy, Clark.
Conflict of Interest Disclosures: None reported.
Funding/Support: The Walker Premature Infant Fund at the Baylor Health Care System Foundation and the Division of Nursing at Baylor University Medical Center provided funding for the statistical analyses performed.
Role of the Funder/Sponsor: The sponsors had no involvement 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.
Additional Contributions: Arpitha Chiruvolu, MD, Pediatrix Medical Group, Baylor Health Care System, reviewed the manuscript. Dr Chiruvolu did not receive financial compensation.
Correction: This article was corrected on November 3, 2014 for a typographical error in the text.