Palivizumab and Long-term Outcomes in Cystic Fibrosis

This retrospective study was conducted by using longitudinal CFFPR data from 2008 to 2016. The study start date (January 1, 2008) was the first year that the CFFPR included data on palivizumab receipt. The CFFPR includes individuals with CF who receive care at a CFF Care Center network–accredited program. The CFFPR population, data collection methods, validity, and generalizability were previously described by Knapp et al.¹⁶ 

Infants eligible for this analysis were born between January 1, 2008, and December 31, 2015; were diagnosed with CF within the first 6 months of life; and had ≥1 calendar year of CFFPR data during their first 2 years of life. Infants without annual data during the first 2 years of life or without encounter data recorded before their first RSV season were excluded. Each participant was followed through December 31, 2016, or December 31st of the year in which they turned 7 years old. The Quorum Review Institutional Review Board approved this study (32599/1).

The exposure of interest was palivizumab receipt during the first 2 years of life. In this analysis, “age” describes age on December 31st of the calendar year. Infants were categorized as receiving palivizumab if during either or both of their first 2 years of life it was recorded in the registry. The CFFPR does not record the number and timing of doses.¹⁷ Infants were categorized as not receiving palivizumab if during either or both of their first 2 years of life palivizumab was recorded as “not given.” Sensitivity analyses were performed, removing infants with 1 year when palivizumab was recorded as not given and palivizumab receipt was unknown for the second year. Infants were excluded if this field was not completed for both of their first 2 years of life.

Three longitudinal outcomes were assessed: (1) annualized percent FEV₁ predicted (average of the maximum measure from each quarter) during the 12 months before the seventh birthday,¹⁸ (2) time to first positive P aeruginosa culture result during the first 7 years of life, and (3) age-specific annual number of hospitalizations for pulmonary exacerbations or other pulmonary complications during the first 8 years of life. Lung function measures from between their sixth and seventh birthdays ensured participants were old enough to provide accurate pulmonary function measures. Percent predicted was calculated by using Global Lung Function Initiative reference equations. Infants with a positive culture result for P aeruginosa before palivizumab receipt were excluded from the analysis of time to first positive culture result for P aeruginosa.

Potential confounders of the association between palivizumab receipt and outcomes were explored. Demographic variables included year of birth, birth season (January–March, April–June, July–September, or October–December), sex, race, and ethnicity. Gestational age was categorized as ≥37, 32–<37, or <32 weeks or missing. CF diagnosis–related variables included cystic fibrosis transmembrane regulator (CFTR) mutation class (classes I, II, or III; classes IV and V; unclassifiable; or not genotyped)¹⁹ and method of diagnosis and/or symptoms at diagnosis. Individuals with 2 class I to III mutations are typically pancreatic insufficient with more severe lung disease, whereas those with ≥1 class IV or V mutation are typically pancreatic sufficient with milder lung disease. Symptoms at diagnosis were classified by using hierarchical categories: meconium ileus, respiratory symptoms, newborn screening, and other (eg, gastrointestinal symptoms or failure to thrive). Additional annual variables were categorized as present if recorded “yes” and absent if recorded “no” during the first 2 years of life. These variables included a smoker in the household, day care attendance, influenza vaccine receipt, and Medicaid. Household size, including the infant, was obtained during infants’ first years in the CFFPR and was categorized as 2 to 3 individuals, 4 to 6 individuals, and ≥7 individuals. Many variables collected annually had missing data; to minimize excluding otherwise eligible infants, a “missing” category was created for each annual variable.

We addressed the possibility that sicker infants were more likely to receive palivizumab (confounding by indication) by assessing clinical status during the time that clinicians and families considered palivizumab. We examined clinical status before first RSV season using CFFPR measures available before infants’ first November first, including number of clinic visits; last recorded length, weight, and weight-for-length percentiles²⁰; Staphylococcus aureus, methicillin-resistant S aureus, Haemophilus influenzae, and P aeruginosa detected on any respiratory culture (yes or no for each); use of inhaled tobramycin, dornase alfa, hypertonic saline, and/or pancreatic enzyme replacement therapy (yes or no); and reported asthma in the CFFPR (yes or no).²¹ Infants without clinical encounters from birth to their first RSV season were excluded.

Bivariate analyses used χ² tests for categorical variables and t tests for continuous variables. We created propensity scores to adjust for potential confounding by indication, calculated using logistic regression models to predict the probability of receiving palivizumab. To create the propensity model, we first created models to determine variables that were associated with each outcome of interest. All variables included in those models with P < .1 were included in the final propensity model.

For each outcome of interest, we compared those who did and did not receive palivizumab. We performed unadjusted regression analyses, then adjusted for the propensity score. Multivariate linear regression was used to estimate difference in percent FEV₁ predicted. Kaplan-Meier curves and Cox proportional hazards models were used to estimate timing and hazard of first P aeruginosa acquisition. Lastly, negative binomial Poisson models were used to model annual count of hospitalizations for pulmonary complications and generate rate ratios (RRs).

We performed sensitivity analyses of population subsets to compare this analysis with other literature on palivizumab and to reflect subgroups indicated in the AAP recommendations. We examined risk of death ever in the first 2 years of life and specifically during the RSV season. We also restricted the analysis of hospitalization to the first 2 years of life and focused specifically on those occurring during RSV season. Lastly, we analyzed all outcomes of interest only among infants categorized as high risk, defined as either having a weight or weight-for-length percentile of <5% or having a hospitalization for a pulmonary exacerbation or other pulmonary complications before their first RSV season.

Analyses were conducted by using SAS version 9.4 (SAS Institute, Inc, Cary, NC).

Between 2008 and 2015, 6032 infants were diagnosed with CF within the first 6 months of life. From these, we excluded 516 with missing data for palivizumab receipt in both years, 333 infants without annualized data in the first 2 years of life, and 916 infants without clinical encounters before their first RSV season. Therefore, the study sample included 4267 infants, with 1588 (37%) reported as receiving palivizumab in the CFFPR during ≥1 of the first 2 years of life. Of these, 31% received palivizumab in both years, 38% in year 1 only, 10% in year 2 only, 9% in year 1 and unknown in year 2, and 12% in year 2 and unknown in year 1.

Among the 2679 children reported as not receiving palivizumab, 971 (36%) had 1 year in which palivizumab was reported as not given but another year in which palivizumab receipt was reported as “unknown” or was missing. The percentage of infants at each care site reported as receiving palivizumab varied (median: 37%; interquartile range [IQR]: 18%–64%).

Table 1 reveals study population demographics and clinical characteristics. Infants receiving palivizumab were more likely to be born in 2008–2012, born between April and September, have a <37-week gestational age, have 2 CFTR class I to III mutations, and meconium ileus at birth. Before their first November first, infants receiving palivizumab had lower height and weight percentiles; were more likely to have been prescribed dornase alfa, hypertonic saline, and pancreatic enzyme replacement therapy; and were more likely to have asthma reported in the CFFPR.

Variables included in the propensity model are noted in the Table 1 footnote. There was considerable overlap of calculated propensity scores. Infants who received palivizumab had a median propensity score of 0.41 (IQR: 0.33–0.48), and infants who did not receive palivizumab had a median propensity score of 0.35 (IQR: 0.28–0.42).

Of the 4267 cohort participants, 1323 (31%) had lung function data recorded in the CFFPR between ages 6 and 7; 593 of these children had received palivizumab and 730 had not. Mean percent FEV₁ predicted at 7 years of age was similar among those who received palivizumab (98.2; 95% confidence interval [CI]: 96.9–99.5) and did not receive palivizumab (97.3; 95% CI: 96.1–98.5), adjusting for propensity scores. The sensitivity analysis that removed infants who had 1 year in which palivizumab was recorded as not given and palivizumab status was unknown for the other year did not change these results (data not shown).

There were 3619 infants who had negative P aeruginosa culture results reported in the CFFPR at the beginning of their first RSV season (1340 had received palivizumab and 2279 had not) and thus were included in analysis of time to first P aeruginosa–positive culture result. During follow-up, 1865 children (52%) had P aeruginosa–positive culture results. Unadjusted Kaplan-Meier curves in Fig 1A suggest those who received palivizumab had significantly greater risk of acquiring P aeruginosa (log-rank test, P = .007). After stratification by propensity score quartiles, the differences were no longer significant (Quartile 1, lowest propensity to receive palivizumab P = .8; Quartile 2, P = .2; Quartile 3, P = .2; Quartile 4, highest propensity to receive palivizumab P = .8). Figure 1B reveals Kaplan-Meier curves for infants in the quartile with lowest propensity to receive palivizumab and Fig 2 shows curves for those with highest propensity. The unadjusted hazard ratio for acquisition of P aeruginosa was 1.1 (95% CI: 1.0–1.2) and after propensity score adjustment, the hazard ratio was 1.1 (95% CI: 0.96–1.2). Estimate of effect did not change when we restricted analysis to unexposed infants whose palivizumab status was known for both years (data not shown).

Table 2 reveals RRs for hospitalizations by age. During follow-up at any age, the percent of children with ≥1 hospitalization each year ranged from 13% to 20%. The adjusted RR was highest between ages 1 and 2 (RR: 1.1; 95% CI: 0.92–1.3) and lowest between ages 7 and 8 (RR: 0.75; 95% CI: 0.54–1.0). Overall, unadjusted RR were attenuated after adjustment for propensity score.

We conducted numerous sensitivity analyses. Unadjusted risk of death before age 2 was 0.1% among infants who did and 0.2% among those who did not receive palivizumab (P = .5). Restricting to deaths during the RSV season, risk was the same (0.1%) among infants treated and not treated with palivizumab (P = .6). Small sample sizes precluded additional analyses. RR for hospitalizations before age 2 was 1.0 (95% CI: 0.89–1.2) comparing those who received and did not receive palivizumab. Restricting to hospitalizations during RSV season, the RR was 0.98 (95% CI: 0.84–1.1).

Overall, 1426 infants were categorized as high risk, 604 (38%) of those treated with palivizumab and 822 (31%) of infants not treated with palivizumab. Restricting analyses to these high-risk infants, 462 infants were included in the lung function analysis. Mean percent FEV₁ predicted was the same in those who received and did not receive palivizumab (95.3 vs 95.0; P = .9). The hazard ratio for acquisition of P aeruginosa was 1.0 (95% CI: 0.86–1.2) among 1103 high-risk infants, comparing those who did versus did not receive palivizumab. Table 2 reveals RRs restricted to high-risk infants, suggesting potential for decreased hospitalizations in younger ages.

To our knowledge, this is the first large population-based study evaluating association of palivizumab receipt in infancy with long-term outcomes in children with CF using propensity scores to minimize indication bias. As expected, infants receiving palivizumab were more likely to be born preterm and before RSV season; have severe CFTR mutations; be diagnosed because of meconium ileus; be diagnosed with asthma; and be treated with dornase alfa, hypertonic saline, and pancreatic enzymes. These factors were likely considered by providers and families when prescribing palivizumab and are somewhat consistent with AAP recommendations for palivizumab in infants with CF with chronic lung disease and/or nutritional compromise. After adjustment for propensity scores, we found no association between palivizumab and lung function at age 7, time to first P aeruginosa–positive culture result, or rate of pulmonary-related hospitalizations.

Authors of other observational studies evaluating long-term outcomes of palivizumab did not overall find impact of palivizumab on these outcomes. In Northern Ireland, investigators examined time to first P aeruginosa, lung function, and growth parameters in children with CF born before (n = 47) and after (n = 45) introduction of palivizumab in 2002.⁹ Like our study, lung function and growth parameters were similar; however, unlike our study, the “after” group had a significantly shorter time until first P aeruginosa–positive culture result compared with the “before” group (median: 57 vs 96 months, respectively). In France, investigators examined the acquisition of selected microorganisms in infants who had (n = 40) or had not (n = 140) received palivizumab.¹³ Among infants who had and had not received palivizumab, age at first positive culture result was similar for P aeruginosa (10.4 vs 12.3 months, respectively) and for S aureus (6.4 vs 3.8 months, respectively). Those who had received palivizumab were more likely to be positive for S aureus at age 3 than those who had not (97% vs 85%, respectively). Growth outcome measures were similar in both groups throughout the first 3 years of life.

Authors of 1 randomized placebo-controlled study assessed outcomes associated with palivizumab in infants with CF.⁷ At a 12-month follow-up, proportions of infants with positive culture results for P aeruginosa were similar, as were weight-to-height ratios and rates of RSV hospitalizations.⁵ Authors of this study also examined safety. Overall adverse event (AE) incidence was similar: 86 of 92 (93%) infants in the palivizumab group and 90 of 94 (96%) in the placebo group had a reported AE, and 19 of 92 (21%) infants in the palivizumab group and 16 of 94 (17%) in the placebo group experienced a serious AE. Although an adequately powered, randomized controlled trial is ideal for studying the safety and efficacy of palivizumab in infants with CF, many have concluded that the sample size needed to establish safety and efficacy of palivizumab in infants with CF would be too large to be feasible.

Authors of more recent large population-based studies provide evidence that RSV is a risk factor for hospitalization in infants with CF. In Denmark, risk and severity of RSV hospitalizations were assessed in infants <2 years of age with chronic conditions, including CF.²² RSV-related hospitalizations were responsible for 18% of all hospitalizations, representing a 4.32-fold increased risk of RSV hospitalization compared with otherwise healthy children. During the study period (1997–2003), RSV was detected by antigen, which has a lower sensitivity than the assays currently in widespread use today, suggesting underestimation of RSV-related hospitalizations. Similarly, in England, among 11 infants with CF hospitalized with RSV and/or bronchiolitis during a study conducted from April 2007 to March 2008, the rate of admission was 2.66-fold higher in infants with CF than otherwise healthy term infants.²³ 

Nonetheless, lack of quality evidence for benefit, high cost, and operationalizing the complex monthly administration of palivizumab has led to variability in use in the United States, United Kingdom, and Canada.^10,11,24 We noted similar variability in the US CF care center network, as the median percentage of infants receiving palivizumab at individual care centers was 37% (IQR: 18%–64%). Furthermore, infants born from 2008 to 2012 were more likely to have received palivizumab than infants born from 2013 to 2015, which may reflect growing uncertainty about the potential benefits of palivizumab and/or more barriers restricting access to this therapy.

This registry-based observational study has several limitations. In this large cohort of infants with CF, 29% of the potential study sample was excluded because of missing data, potentially impacting generalizability of results. Approximately 81% to 84% of people with CF in the United States are included in the CFFPR, but the precise proportion of infants and children included in the CFFPR is unknown.¹⁶ Infants with missing data on palivizumab receipt were more likely to be born in recent years, be nonwhite, and have low gestational age. There was likely some misclassification of palivizumab exposure status, which could have biased our results toward the null. Palivizumab prescribed by primary care or other non-CF care center clinicians may not be recorded accurately in the CFFPR. Number of palivizumab doses received in each season, which can potentially impact effectiveness, is not captured in the CFFPR. The CFFPR does not contain RSV disease data, and thus hospitalizations secondary to RSV could not be assessed during treatment, the traditional benefit of treatment. Our propensity scores may not have adjusted completely for indication bias and could only include measured confounders. Finally, pulmonary function data were only available for a study population subset because most had not turned 7 years old by the end of the observation period.

We found no evidence of long-term benefits associated with palivizumab in infants with CF on a population level. The fact that, after adjustment for indication bias, those who did or did not receive palivizumab had similar long-term outcomes could be seen as reassuring. However, current AAP guidelines recommend palivizumab for high-risk infants with CF (ie, those with chronic lung disease and nutritional compromise). An appropriate next step to evaluate outcomes of palivizumab among high-risk infants would be to embed a prospective study within the CFFPR. First, concise case definitions for chronic lung disease and nutritional compromise for infants with CF need to be developed. Then, outcomes would be tracked by using current CFFPR data. Additionally, accuracy of palivizumab receipt, dose timing, RSV hospitalizations, and hospital and ICU days could be collected by linking the CFFPR to administrative claims data.

We thank the CFF for the use of CFFPR data to conduct this study. In addition, we thank the patients, care providers, and clinic coordinators at CF centers throughout the United States for their contributions to the CFFPR.

AAP

American Academy of Pediatrics

AE

adverse event

CF

cystic fibrosis

CFF

Cystic Fibrosis Foundation

CFFPR

Cystic Fibrosis Foundation Patient Registry

CFTR

cystic fibrosis transmembrane regulator

CI

confidence interval

FEV₁

forced expiratory volume in 1 second

IQR

interquartile range

RR

rate ratio

RSV

respiratory syncytial virus