Factors predicting future ACS episodes in children with sickle cell anemia

Introduction

Asthma is associated with an increased rate of pain and acute chest syndrome (ACS) in sickle cell anemia (SCA). However, the diagnosis of asthma in SCA is particularly challenging because clinical features commonly associated with asthma, such as wheezing 1, an elevated IgE level 2, increased response to bronchodilators 3, 4, and airway hyper-responsiveness 4-6 can occur in children with SCA with or without a diagnosis of asthma.

Well-established challenges are present in making a consistent diagnosis of asthma in children 7. Thus, as part of the primary aim of the NHLBI funded prospective observational cohort study in SCA, the Sickle Asthma Cohort (SAC) study, we tested the following hypothesis: in children with SCA, respiratory symptoms and other risk factors for asthma are associated with an increased incidence of hospitalizations for pain and ACS. If criteria, other than a physician diagnosis of asthma 8, or early onset of ACS 9, can be identified, these risk factors could be used as prognostic markers for children at greater risk for severe pain episodes and ACS events. This would facilitate stratification of management targeting those features that increase risks of ACS or pain.

Methods

Institutional approval was obtained from participating sites, Washington University School of Medicine in St. Louis, Missouri; Case Western Reserve in Cleveland, Ohio; and University College London in London, United Kingdom. Three pediatric hematology centers participated in London via the National Health Service Research Ethics Committee approval, Guys and St. Thomas NHS Foundation Trust, Imperial College and Imperial College Healthcare NHS Foundation Trust and North Middlesex University Hospital NHS Trust. Institutional approval was also obtained from the Coordinating Center at Vanderbilt University School of Medicine in Nashville Tennessee.

In 2005, we initiated our prospective observational cohort study of children age 4–18 years with SCA (Hgb SS [homozygous for the β^s-globin mutation] or sickle-betaº-thalassemia [Hgb Sβº]). Children were excluded from enrolling in the observation study if they: (i) were on regular blood transfusion therapy; (ii) in a clinical trial evaluating hydroxyurea or oxygen therapy; or (iii) had any additional co-morbidity associated with lung disease other than asthma.

Definition of outcomes

A vaso-occlusive pain episode was defined as an episode directly associated with SCA, which required hospitalization, and was treated with opioids. Headaches that required admission to the hospital and were treated with opioids were not considered a vaso-occlusive pain episode. Acute chest syndrome (ACS) was defined as an episode of acute respiratory distress associated with at least one new radiodensity on chest roentgenogram, temperature >38°C and increased respiratory effort with decreased oxygen saturation or increased respiratory rate documented in the medical record. A diagnosis of pneumonia was considered an ACS episode.

Clinical Features Associated with Asthma, Atopy, Lung Function, and Respiratory Symptoms. Allergy skin tests were performed by technicians who were centrally trained and certified using the prick puncture technique (MulitTestII device, Lincoln Diagnostics, Decatur, IL) to nine aeroallergens (Aspergillus and Alternaria molds, cat, dog, dust mite, and cockroach, and site specific tree, grass, and weed pollens). Each wheal was outlined with a skin-marking pen. Each outline was transferred with transparent tape to the boxed areas on the skin test results form. After removing the tape from the child's back or arm, a Paper Mate Gel Writer pen was used to draw line “a”, which represented the longest length of the wheal. Line “b” was drawn perpendicular to “a” at the widest width. Line “b” was not necessarily the midpoint of line “a”. The length of both lines were taken inside the ink outline. These measurements in millimeters were recorded on a skin test result sheet. A testing session was considered valid if the histamine positive control had a mean wheal size of at least 3 mm greater than the measurement from the saline control. The result of an individual test for an aeroallergen was considered positive if the mean wheal size of at least 3 mm greater than the saline control. Results were re-measured by a single investigator to ensure reliability. Two or more positive tests versus less than 2 positive tests were considered positive results. While atopy is often defined as 1 positive result, two positive skin tests were used as the indicator of atopy to give assurance of clear evidence of atopic status, as data from the Childhood Asthma Management Program research group has found that more than 1 positive skin test is better correlated with methacholine responsiveness 10, hospitalization 11, and nocturnal awakenings 12.

Spirometry was performed before and after bronchodilator, administered as four inhalations of albuterol, 90 mcg/inhalation, via an aerochamber; all results were over-read by a single certified pulmonary function technologist using ATS criteria appropriately modified for children 13 with only those meeting these criteria being used in subsequent analyses. Spirometric results were adjusted for age, sex, height, and ethnicity using the GLI-2012 all age-multi-ethnic reference equations for Black subjects 14. Airway obstruction was defined as the forced expired volume in 1 second/forced vital capacity (FEV₁/FVC) < lower limit of normal (LLN) according to these equations 14 and a significant bronchodilator response was defined as an increase in FEV₁ ≥ 12%.

During the initial evaluation, all parents of the participants were asked to complete the American Thoracic Society Division of Lung Disease (ATS/DLD) questionnaire to ascertain the presence of respiratory symptoms 15.

Data quality

To ensure a uniform definition of pain and ACS episodes in this multi-center study, the charts of all patients diagnosed with ACS or a vaso-occlusive pain episode requiring hospitalization for pain in the chest, extremities or other areas of the body, were reviewed by a single investigator at each of the participating sites after training by the PI, who reviewed all the hospital visits at St. Louis and the de-identified clinical documentation for 100% of hospital visits for the first four years of the study at the Cleveland and London sites. When concerns about the assignment of the diagnosis were raised, clarity of the assessment was discussed with the site investigators.

Construction of SAC cohort

All patients included in the analysis had been followed from birth and for at least 3 years subsequent to enrolling in the study at one of the three sites. Follow-up from birth was deemed necessary because we have previously demonstrated that an ACS episode prior to 4 years of age was associated with a doctor diagnosis of asthma 16. Specifically for this analysis, we wanted to determine whether an early event of pain or ACS was associated with future ACS and pain events. Figure 1 describes the flow diagram for study participants who were included in these analyses. All patients in the cohort were prospectively followed from the date of informed consent until death, start of regular blood transfusion therapy, or last date of follow-up, whichever came first.

Validation of cohort findings obtained in the SAC study

Data from the Cooperative Study for Sickle Cell Disease (CSSCD) infant cohort data set were used to confirm the primary results of the study. For all practical purposes, the infant CSSCD cohort was a birth cohort due to the average age of enrollment of <6 months, so we modeled the prospective pain and ACS events after 4 years of age to match the age of entry of the SAC study participants. An ACS episode was defined as a new pulmonary infiltrate demonstrable on chest radiograph. Figure 2 describes the flow diagram of the CSSCD study participants selected for this analysis. In the CSSCD cohort, unlike the SAC cohort, asthma predictors excluded spirometry evaluation and the ATS/DLD questionnaire.

Statistical analysis

Analyses were conducted using Stata statistical software (Version 12, College Station, TX: StataCorp LP) and IBM SPSS Statistics (Version 22, Chicago, IL, IBM). Continuous data were analyzed using t-tests for variables with a normal distribution and the Wilcoxon–Mann–Whitney test for variables with non-normal distributions. Categorical data were analyzed using chi-square tests. Rates of missing data for the primary analysis in each covariate ranged from 0 to 3.0%. Multivariable negative binomial regressions were performed using characteristics postulated to be relevant to predicting rates of pain or ACS, including the previously documented association between an early episode of ACS and subsequent risk of ACS 9. Regression was conducted in two steps. First, a model was constructed with all the relevant predictors and a significance of P < 0.20 was used to screen for inclusion in a reduced model. Then a second model was run with the screened set of characteristics. Collinearity was assessed with a criterion variance inflation factor < 3.0.

Duration of hydroxyurea therapy ranged from 0.10 to 13.92 years (mean 3.59 years, median 2.50 years). The incidence rates of pain or ACS were not decreased by use of hydroxyurea (n = 52) during the study period (Table 1), we therefore did not include hydroxyurea medication as a covariate in the multivariable analysis.

Table 1. Pain and ACS Rates for Time on and off Hydroxyurea in the SAC Cohorta

Type of event	On hydroxyurea	Off hydroxyurea	P-valueb
Pain rate (events/patient year)	1.02 (2.41)c	0.56 (0.99)	0.003
ACS rate (events/patient year)	0.18 (0.71)	0.25 (0.40)	0.891

• Duration of hydroxyurea therapy ranges from 0.10 to 13.92 years. The mean is 3.59 years, median is 2.50 years.
• ^a Includes only those followed since birth and administered hydroxyurea; n = 52.
• ^b Wilcoxon signed rank test.
• ^c Median (IQR).

The measures of spirometry evaluation and skin testing were added to the model with asthma symptoms and clinical risk factors in a separate regression. We use this modeling strategy because it more closely mimics clinical practice, as clinicians would be expected to take a history and ask a series of questions about asthma symptoms and clinical risk factors before ordering tests to evaluate the presence of additional asthma risk factors (spirometry results and skin testing).

The minimum age of the cohort eligibility, 4 years, decreased the number of children with evaluable spirometry (n = 144 of 159 possible participants). Some members of our cohort do not have skin test data (n = 18) because several parents did not agree to the skin testing and some patients had negative histamine controls making their results uninterpretable. Consequently, the expanded regression model included 130 participants with complete data on bronchodilator response, skin test, and lower airway obstruction and not the 159 participants in the asthma symptoms regression model. There were few differences between those patients with missing data who were excluded from the expanded model and those with complete data (Table 2).

Table 2. Comparison of Patients in the SAC Study Missing Data on Covariates in Models Including Positive Skin Tests and Spirometry to those Patients with Complete Data

Characteristic	No missing data	Missing data	P-valuea
	(n = 130)	(n = 29)
Sex (% males)	53.8	41.4	0.224
Age (yrs)c	10.00	8.00	0.031
Prospective follow-upb (yrs)c	5.46	4.49	0.004
Pain rate, lifetime (events/patient year)c	0.38	0.30	0.623
Pain rate, prospective (events/patient year)c	0.44	0.50	0.862
Pain event before age 4 years old, (%)	53.1	51.7	0.895
ACS rate, lifetime (events/patient year)c	0.16	0.08	0.041
ACS rate, prospective (events/patient year)c	0.17	0.00	0.363
ACS event before age 4, (%)	42.3	20.7	0.030
Hemoglobin (g/dl)d	8.21	8.41	0.440
White blood cells, 109/ld	12.13	12.09	0.962
Mother has asthma, (%)	11.5	20.7	0.188
Wheezing cause shortness of breath, (%)	25.4	31.0	0.533
Wheezing after playing hard, (%)	35.4	24.1	0.245
Wheezing without having a cold, (%)	22.3	13.8	0.307
Cough without having a cold, (%)	36.2	44.8	0.384
Congested or phlegm without a cold, (%)	13.1	17.2	0.557
Two or more positive skin tests, (%)e	27.7	18.2	0.727
Bronchodilator response ≥ 12%, (%)e	22.3	14.3	0.735
Lower airway obstruction, (%)e	20.8	7.1	0.305

• ^a Chi-square test for categorical variables; t-test for variables with mean reported; Mann–Whitney U test for variables with median reported.
• ^b Prospective follow-up is from date of entry in SAC; Minimum follow-up of 3 years.
• ^c Median (IQR) for continuous variables with a non-normal distribution.
• ^d Mean (std dev.) for continuous variables with a normal distribution.
• ^e Eighteen cases with data in missing group for two or more positive skin tests; 15 cases in missing group for bronchodilator response and lower airway obstruction, respectively.

Results

Discussion

A doctor-diagnosis of asthma in children with sickle cell disease (SCD) is associated with increased SCD related morbidity 2, 8, 16, 17 and mortality 18. In the SAC cohort specifically, asthma diagnosis was associated statistically with ACS occurring after entry into the study 8. However, no comprehensive evaluations of specific asthma signs and symptoms have been conducted to determine whether they are associated with SCD morbidity. A priori, we elected not to include a physician diagnosis of asthma in our study design because not only has the diagnosis of asthma in children with SCA proven to be challenging given the overlap in asthma symptoms that are common in children with SCA that do not have asthma 19, 20, but there is considerable variability in general in making an asthma diagnosis 7. We instead focused our primary analysis on asthma symptoms and risk factors, which we believed would make our results more generalizable. We elected to include ACS events before 4 years of age in our analyses, as Boyd et al. demonstrated that those with asthma were younger at the time of the first ACS episode than those without an asthma diagnosis 16 and Quinn et al. found that early ACS events predicted subsequent ACS events 9. The results of our prospective study indicate that at least one early life ACS episode was a significant predictor of later ACS events, and this finding was replicated in a second cohort.

One asthma symptom and one asthma risk factor were associated with future ACS events, specifically history of wheeze causing shortness of breath and presence of two or more positive allergy skin tests. As the children were studied when well and at least one month since a respiratory illness, it is not surprising that airway obstruction or bronchodilator response were not significant covariates in the final multiple variable models.

The mechanism for the association of wheezing and aeroallergen sensitivity to future ACS events is unknown, but two experimental studies undertaken in the transgenic SCD mouse models and two clinical studies completed in children by our team provide a basis to postulate. First, transgenic SCD mice have increased airway lability when compared to hemoglobin A mice 21 at baseline. Second, when sensitized to an inflammatory stimulus, ovalbumin, SCD mice have an additionally exaggerated inflammatory airway response, as demonstrated by increased levels of peri-alveolar eosinophils, plasma IgE, and bronchoalveolar lavage fluid IL-1β, IL-4, IL-6, and monocyte chemotactic protein, when compared to the hemoglobin A mice 21. Supporting the findings that children with SCA have an increased sensitivity to inflammatory stimuli, two separate studies documented that in children with SCA, total and specific IgE levels, were dramatically elevated when compared to expected ranges in children without SCA 2, 22.

Based on pre-clinical and clinical studies demonstrating a propensity for increased inflammatory response to stimuli, we postulate that similar to the SCD mouse model, external inflammatory stimuli result in an exaggerated response in children with SCD and lead to increased pre-disposition to ACS. For children less than 4 years of age, the age group most susceptible to multiple viral respiratory infections, we posit that these acute respiratory viral infections predispose the young child to development of increased airway inflammation in the near-term and asthma in the long-term. These changes would then predispose to additional SCD events in the near- and long-term. Only future basic science and translational research in this area will ultimately confirm whether acute viral respiratory infections predispose children to future lung disease. The well-established evidence regarding acute infections of respiratory syncytial virus 23-25 or acute rhinovirus infections 26-28 and their associated predisposition for future obstructive lung disease, coupled with current results, makes the study of early viral infection, an important area for investigating the etiology and potential treatment of lung disease in SCA.

We did not identify an association between asthma symptoms or asthma risk factors and pain; although, all of the effects were in the expected direction in the initial logistic model. The lack of a significant association may be due to either a true or false negative association. The false negative association might be related to the relatively small number of patient-years of follow-up, particularly when compared to other studies that have showed an association 2, 16. Evidence to support this possibility is the observation that the current study had <1,000 patient years accumulated in the analysis, while the two largest cohort studies had approximately 3,000 2 and 4,000 patient years 16, respectively, with both demonstrating a statistically significant relationship between an asthma diagnosis and an increased incidence of both ACS and pain episodes when compared to children without an asthma diagnosis. Based on these studies, asthma symptoms and asthma risk factors are more strongly related to the increased incidence of ACS than pain events.

This study had several limitations. The SAC cohort consisted of participants from three academic medical centers with a high level of SCD and asthma expertise. Thus the patient population cannot be considered generalizable to all children with SCA. However, there was no pre-selection of children at the three sites for lung disease, the prevalence of asthma (28% in the SAC cohort used in the models) was similar to the figure of 27% (140 of 521) in the largest study to date of children with SCA to assess the prevalence of asthma 2. Furthermore, our primary results were replicated in a cohort followed from early infancy where children were not selected for severity of disease 29 and in addition are similar to findings in another cohort followed from birth 9. In the CSSCD cohort, we were unable to confirm whether affirmative answers to the ATS/DLD questions were associated with future ACS or pain episodes, as these data were not collected. However, a history of recurrent severe wheezing were associated with future ACS events in a prospective study of adults 30 and the presence of wheezing was associated with an increased incidence of pain and ACS presentation to the emergency department in a retrospective study in children and adults with SCA 31. Taken together the results from this study and the other two studies indicate that history of wheezing is a significant predictor for ACS in children and adults with SCA.

Based on the results of this study, health care providers can now begin to identify children with SCA at risk for future ACS events with simple questions about the presence of wheezing causing shortness of breath coupled with obtaining a history of an ACS episode prior to 4 years of age. Future research is needed to better understand the reasons underlying the peak incidence of ACS in SCA in children between 1 and 4 years of age, and how this leads to an increased incidence rate of future ACS events as clearly documented in this study.

Acknowledgments

This work was supported by the NHLBI, R01-HL079937, MRD, and RCS.