Introduction

Sickle cell disease (SCD) is the most common monogenic disease in the United States, resulting from mutations in the beta-subunit of the hemoglobin molecule. Acute episodes of pain or vaso-occlusive crises (VOCs) are a protean hallmark of SCD with economic impact due to the cost of unscheduled health care[1]. Acute VOCs are due to erythrocyte microvascular occlusion and tissue hypoxia, especially in sickle cell anemia (SCA)[2]. Inter-individual differences in severe VOCs leading to utilization of acute care are more common in a sub-group with frequent pain[3–6]. Reasons for this variability have not been elucidated, but higher hematocrit and lower fetal hemoglobin (HbF) are strong predictors for frequent VOCs[4]. Frequent VOCs defined by physician treatment and a duration of more than 2 hours were associated with higher mortality in the Cooperative Study of Sickle Cell Disease (CSSCD) although this association has not been evaluated in contemporary populations where health care utilization for acute painful episodes remains common despite the availability of SCA specific treatments[3,4,7].

SCA is characterized by specific complications that are risk factors for early death, although modern therapies are likely to have altered the natural history and mortality of SCA[8–10]. In particular, hydroxyurea (HU) in SCA is an effective medication to significantly reduce hospitalizations for pain[9,11]. Scheduled chronic transfusion in children at high risk for stroke prevents both cerebrovascular events and reduces the prevalence of VOCs[8,12]. Additionally, an elevated tricuspid regurgitant velocity (TRV) is a risk factor for death in some series and suggestive of cardiopulmonary disease, although trials of pulmonary vasodilators have failed to demonstrate a clinical benefit[13–17]. Based upon these observations, we hypothesized that therapies with demonstrated SCA efficacy like transfusion programs and HU have altered the magnitude of associations between acute painful episodes, health care utilization and mortality. We sought to 1) identify markers associated with VOCs for comparison to established VOC associations with baseline hemoglobin and HbF expression and 2) determine if acute painful episodes defined by hospital based treatment as a patient reported outcome (PRO) are a relevant measure of disease severity and risk for death in a contemporary SCA cohort at a single referral institution.

Methods

Results

Patient reported severe pain crises are an independent risk factor for mortality in adults with SCA

To determine if severe pain crises remain a risk factor for early mortality in contemporary SCA patients, univariate survival analysis was performed. During a median follow-up of 4.92 years, 40 (15.1%) subjects died and 9 (3.4%) were lost to follow-up. Kaplan-Meier showed a younger age at death in the high VOC group (55.8 vs. 66.2 years; P=0.04) (Figure 1). Kaplan-Meier also showed a significant difference when comparing 3 pain sub-groups (0 ED visits/hospitalizations, 1-4 ED visits/hopitalizations and ≥5 ED visits/hospitalizations) (P=0.02 for trend, Figure S1). Because the Kaplan Meier curves comparing the 2 pain groups (1-4 versus ≥5 ED visits/hospitalizations) were not significantly different (P=0.22, Figure S1), a Cox proportional hazards analysis of mortality utilized the categorical high (≥1 ED visits/hospitalizations) versus low VOC group classification. The risk of death among 208 evaluable subjects with 35 deaths during follow-up by proportional hazards modeling was again significantly higher with more than 1 severe VOC during the past 12 months (relative risk [RR] 2.68, P= 0.03), a higher TRV (RR 2.41, P<0.0001), elevated ferritin (RR 4.00, P=0.001) and low estimated glomerular filtration rate (RR 2.73, P<0.0001) (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Kaplan Meier (KM) curve showing survival in sickle cell anemia by severe pain crises requiring an ED visit/ hospitalization in past year.

https://doi.org/10.1371/journal.pone.0079923.g001

Risk factor	RR (95% CI)*	p Value†
ED/Hospitalization, ≥1 in past year	2.68 (1.1-6.5)	0.03
TRV‡	2.41 (1.5-3.9)	<0.0001
Ferritin	4.00 (1.8-9.0)	0.001
GFR	2.73 (1.6-4.6)	<0.0001

Table 3. Cox proportional hazards mortality in adults with SCA.

* RR for ED/hospitalization is given relative to no visits in the past year. RR for TRV, ferritin and GFR are given for the 75^th relative to the 25^th percentile.

^† From Wald χ².

^‡ 21 subjects with undetectable TRV by echocardiogram were excluded from analysis. TRV boundaries for the 75^th and 25^th percentiles were 2.8 m/s and 2.2 m/s, respectively.

Abbreviations: CI, confidence interval; ED, emergency department; GFR, Modification of Diet in Renal Disease estimated glomerular filtration rate; RR, relative risk; SCA, sickle cell anemia; TRV, tricuspid regurgitatant jet velocity.

CSV

Download CSV

Discussion

We examined VOCs with healthcare utilization and their association with mortality in a contemporary SCA cohort with a high prevalence of HU treatment, a therapy that was not available to subjects included in the first report of this association[4]. We also describe clinical and laboratory characteristics associated with patient reported severe VOCs resulting in an ED visit or hospitalization. Although other SCD genotypes (hemoglobin SC and S beta plus thalassemia) experience vaso-occlusive pain, the present analysis was restricted to sickle cell anemia (homozygous hemoglobin S) due to well known differences in laboratory parameters and the frequency of pain between these genotypes[4,7,22]. In our cohort, about 40% of individuals with SCA reported the absence of hospital admissions or acute treatment in an ED for severe VOCs in the 12 months prior to their initial evaluation on this study, while 22 % reported 5 or more ED visits or hospitalizations during the same time period. These findings highlight the variability of severe painful VOCs within a single genotype despite modern availability of SCD specific therapies.

Despite reports of overall improved survival in SCD, VOC remains associated with premature death similar to results of the CSSCD and likely remains a suitable marker of disease severity[4,7]. The association of higher steady state hemoglobin with VOCs again replicates a key observation from the CSSCD, likely due to the effect of higher viscosity in promoting microvascular sludging and vaso-occlusion[23,24]. The association of higher LDH with the no ED visit/hospitalization group shows a strong negative correlation between LDH and hemoglobin, although their effects were not independent of one another by logistic regression. Our finding of higher iron burden in the ED visit/ hospitalization group could reflect the higher number of red cell transfusions reported by this group. Worsening of anemia is a frequently observed complication of acute painful VOCs which is often treated with red cell transfusions in a hospital setting. High ferritin could also be a reflection of a high inflammatory state in this group, although other inflammatory markers were not significantly elevated in this high VOC group. Finally, higher HDL cholesterol was independently associated with more frequent VOCs. Hypocholesterolemia is common in SCD and other anemias and is thought to be a manifestation of increased erythroid proliferation[25]. Formation of erythroid cell membranes requires cholesterol, and thus, elevated serum HDL cholesterol may be a marker of less marrow activity in those with SCA and a higher hematocrit[26,27]. Additional studies will be required to determine the mechanism of hypocholestorolemia in SCA and if this marker is relevant to the pathophysiology of vaso-occlusion.

Contrary to the anticipated association, fetal hemoglobin concentration was not significantly higher in the no ED visit / hospitalization group. The antimetabolite HU is effective for reducing complications of SCA including frequent severe painful VOCs, with one of the proposed mechanisms of action being increased fetal hemoglobin production, which in turn reduces sickle hemoglobin polymerization and vaso-occlusion[9]. The absence of this fetal hemoglobin association could be due to many factors including confounding by HU indication, which is an intrinsic limitation of observational studies. Other potential confounding variables not assessed in this study include adequacy of HU dosing, duration of HU therapy and compliance with its use. However, the absence of an association between higher fetal hemoglobin and fewer VOCs has also been reported in a contemporary pediatric SCA cohort that was not treated with HU[28].

The association between painful VOC ED visits and hospitalizations and mortality as reported by the CSSCD almost 20 years ago still remains significant in the contemporary era of SCD specific therapies. Additionally, our analysis shows that more frequent severe VOCs requiring healthcare utilization are associated with mortality independent of other known risk factors including an elevated TRV, renal insufficiency and suspected iron burden. These factors could reflect independent aspects of severe disease and chronic organ damage in the VOC group. We are unable to comment on the causes of death in these individuals since most of the participants were followed and received care in their local emergency departments or hospitals. Our results suggest that individuals with a history of health care utilization for severe pain crises and the presence of any of above mortality risk factors are at a particularly high risk for death and should be treated aggressively with appropriate therapies.

One limitation of this study was the examination of patient reported outcomes (severe VOCs), which could introduce the bias of self-recall and under or overestimation of the number of actual ED visits/ hospital admissions. Daily pain experienced by ambulatory adults is underestimated and often managed at home; therefore ambulatory pain events are not captured by utilization definitions[6]. The presence of such pattern in our study could lead to underestimation of frequency of severe VOC pain crises in our population. Despite these limitations associated with severe pain crisis as a patient reported outcome, this definition of pain in SCA (i.e. the pain crisis rate) has also been useful as a clinical endpoint in drug trials, as a patient reported component of a forthcoming quality of life measure for adults with SCD (ASCQ-Me or Adult Sickle Cell Quality of Life Measurement Information System) and as a prognostic marker for mortality in the CSSCD[7,9,16,29,30]. The pain crisis rate may also represent a heritable phenotype for studies of genetic modifiers in SCA[31]. To some extent, this validates our use of cost effective, patient reported outcomes for a history of pain to assess SCA disease severity.

Our findings suggest that the above lab studies are biomarkers indicative of SCA clinical severity that contributes to premature mortality in contemporary patients receiving modern therapies. Although these associations might represent epiphenomena of clinical severity, they provide several hypothetical targets for future investigation of interventions to improve longevity in SCA.

Supporting Information

Acknowledgments

The authors would like to thank the study participants, the NIH Clinical Center nursing staff, and Britny Hall. Dr. Lena Diaw is also thanked for critically reviewing this manuscript. Trial Registration: Clinicaltrials.gov identifier: NCT00011648.