Introduction

Acute chest syndrome (ACS) is one of the most common complications of sickle cell disease, with a frequency of 10–15 episodes/100 patient-years. ACS is defined as hypoxemia combined with various respiratory symptoms (chest pain, tachypnea, wheezing, or cough), a new pulmonary infiltrate on the chest X-ray, and a fever [1, 2]. Although ACS is usually self-limited, some episodes progress rapidly to acute respiratory failure and cause substantial morbidity or death. Regional alveolar hypoxia and the resulting pulmonary vasoconstriction lead to the occlusion of pulmonary microvessels by sickle cells and to lung injury [3]. Regional alveolar hypoxia may result from pneumonia, emboli from blood clots or bone marrow, chest-wall splinting from rib infarctions, atelectasis, sedation from narcotic analgesics given to treat non-pulmonary crises, and alveolar hypoventilation.

Therefore, improving alveolar oxygenation and augmenting alveolar ventilation may be key objectives of the treatment of ACS, in addition to ensuring pain relief and giving blood transfusions and antibiotics [4]. Sickle cell entrapment can be reversed by restoring normal blood oxygenation [3]. However, supplemental oxygen alone might fail to improve regional alveolar oxygenation, most notably at sites characterized by severe ventilation–perfusion mismatching [3–5]. In addition, established ACS is characterized by rapid shallow breathing [6] and blood gas measurements usually show normal or high partial pressures of carbon dioxide \( ( {\text{Pa}}_{{{\text{CO}}_{ 2} }} ) \) indicating relative alveolar hypoventilation [4]. Incentive spirometry and positive expiratory pressure prevented pulmonary complications in patients at high risk for ACS [7–9], and bilevel positive airway pressure seemed effective in a retrospective study of children with established ACS [10]. However, there is no clinical information on the feasibility, efficacy, and safety of the administration of noninvasive mechanical ventilation (NIV) in adults with ACS. The objective of this pilot study was to evaluate the hypothesis that NIV reduces the number of patients remaining hypoxemic at day 3 when used in adults who have early-stage ACS.

Patients and methods

Patients

The study was conducted prospectively between November 1998 and February 2002 at the Henri Mondor Teaching Hospital. Consecutive adults with sickle cell disease (SS, SC, or Sβ-thalassemia) and clinical symptoms suggesting early-stage ACS were eligible. Early-stage ACS was defined as acute hypoxemia [partial pressure of oxygen in arterial blood \( ( {\text{Pa}}_{{{\text{O}}_{ 2} }} ) \) <80 mmHg on room air] and at least one of the following criteria: thoracic or extrathoracic pain, pulmonary crackles by physical examination, new pulmonary infiltrate on the chest X-ray. Non-inclusion criteria were age less than 18 years, pregnancy, need for immediate blood transfusion or immediate endotracheal mechanical ventilation, and multiple organ failure.

Study design

We included consecutive outpatients seen at the emergency department or Sickle Cell Disease Center of our hospital with early-stage ACS, as well as consecutive inpatients who developed ACS during their hospitalization. Patients could be included twice in the event of recurrent ACS, provided the first episode resolved completely and was followed by hospital discharge before the onset of the second episode. Included patients were allocated at random to either supplemental oxygen alone (oxygen group) or oxygen and NIV (NIV group). All patients in both groups received standard care for ACS, and remained in the step-down unit for at least 3 days.

Standard care used in both groups

Standard care included bed rest, continuous supplemental oxygen, intravenous fluid (30 mL/kg of 5% glucose, up to 2,000 mL/24 h), folate supplementation, analgesics and blood products transfusion, as needed. Amoxicillin [100 mg/(kg day)] was given in patients anytime an infection was suspected (see Online Resource 1 in the Electronic Supplementary Material).

Study therapy

Oxygen given via a facial mask was titrated to maintain the pulse oximetry value \( ({\text{Sp}}_{{{\text{O}}_{ 2} }} ) \) ≥95% in both groups. Patients assigned to the NIV group received sequential pressure support ventilation, with sequences of 1 h every 4 h during the day, via an \( {\text{ONYX'}}_{{{\text{O}}_{ 2} }} \) ventilator (Puritan Bennett, Tyco Healthcare France, Plaisir, France), using a facial mask with nasal protection. The mask was individually selected and adjusted, according to the facial morphology of the patient, the measurement of leaks, and the comfort. The settings were adjusted to obtain an expired tidal volume of at least 7 mL/kg, providing a pressure support level averaging 12 cmH2O. The end-expiratory pressure was set arbitrarily at 5 cmH2O. Oxygen was administered between the NIV sequences, using the same procedure as in the oxygen group.

Monitoring

Pain severity was assessed every 4 h using a visual analog scale ranging from 0 (no pain) to 10 (extremely severe pain). The management of pain was standardized whether patients were receiving oxygen alone or in combination with NIV (see Online Resource 1 in the Electronic Supplementary Material). The total amount of narcotics was recorded in milligrams of morphine equivalent. In addition, the patients scored their pain during the clinical round, at seven sites (chest, each of the 4 limbs, spine, and skull) each one being scored from 0 to 3; the overall site-specific scores were therefore summed from 0 (no pain at any site) to 21 (worst possible). Blood gas values, hemoglobin level, and markers for hemolysis were recorded daily. Arterial blood gases were performed on room air in both groups. Chest X-rays were performed at inclusion and at least on day 3, and analyzed by four quadrants, classified as: 0, no infiltrate; 1, mild-to-moderate infiltrate; and 2, alveolar opacity. The radiographic score could range from 0 to 8 (all quadrants showing alveolar opacity). The patients rated their satisfaction with oxygen or NIV, including mask/ventilator comfort and pain control, on a scale from 0 to 3 (satisfied). Nurses assessed patients’ compliance with oxygen or NIV, including comfort, tolerance, and willingness, using a scale from 0 to 3 (compliant).

Outcome measures and statistical analysis

The data generated by this prospective, randomized, single-center, open study were analyzed on an intention-to-treat basis. A random allocation sequence was generated by using a random-number table. The investigators enrolled consecutive patients and used opaque, sealed, numbered envelopes to assign each patient to one of the groups. This method ensured concealment of the allocation sequence.

The primary outcome was the number of episodes free of hypoxemia at day 3 (i.e., \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) >80 mmHg on room air), assessed on \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) values on day 3 of the study. To estimate the sample size, we used the results of two previous studies. In a retrospective study by van Agtmael et al. [11] hypoxemia persisted on day 3 in 84% of 81 ACS episodes. In a prospective randomized trial by Bellet et al. [7] incentive spirometry significantly decreased pulmonary complications (1/19 vs. 8/19; P = 0.019) in patients with chest or back pain above the diaphragm and no baseline pulmonary infiltrates. We therefore assumed that 80% of episodes assigned to oxygen would be hypoxemic on day 3 and that NIV might decrease the proportion of episodes with hypoxemia on day 3 by at least 50%. With the α risk at 0.05 and the β risk at 0.1 (90% power), 35 episodes would be required in each group.

Secondary outcomes included comparison of the alveolar–arterial oxygen gradient \( ({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} \) when breathing room air, and pain relief, narcotic, and blood transfusion requirements, patient satisfaction and compliance with the respiratory device, and length of stay. Patients were usually discharged from the step-down unit when the following clinical criteria were met: the absence of pain or a very residual one, a temperature ≤37.8°C, a heart rate ≤100 beats/min, a respiratory rate ≤24 breaths/min, and oxygen saturation ≥90% or arterial oxygen partial pressure ≥60 mmHg. The return to baseline status indicated hospital discharge thereafter, although other social or psychological criteria might have been involved in the decision process leading to hospital discharge.

Results are reported as medians and interquartiles (25–75). Demographics and clinical data were analyzed by using standard descriptive statistics. We used the chi-square test or Fisher’s exact test, as appropriate, for categorical data, and the nonparametric Mann–Whitney U test for continuous variables. Within-group changes were analyzed by using the nonparametric signed rank test and the McNemar’s test for continuous variables and categorical variables, respectively. P values <0.05 were considered statistically significant.

Ethical considerations

The institutional review board (Comité de Protection des Personnes dans la Recherche Biomédicale) of the Henri Mondor Teaching Hospital approved the study protocol. Written informed consent was obtained from all patients or their family or healthcare proxies prior to enrollment.

Results

From November 1998 to February 2002, 76 consecutive patients with 80 episodes of ACS were eligible for inclusion. However, nine patients, each with a single episode, refused to participate in the study. The remaining 67 patients had 71 episodes (Fig. 1).

Fig. 1
figure 1

Consort flow chart. Dagger four patients were included twice. One patient was randomly allocated twice to NIV; another was randomly allocated twice to oxygen. The two other patients were subsequently allocated to NIV then to oxygen

Full size image

Patient characteristics at baseline

Table 1 compares the main baseline characteristics in the two groups. Patients (37 females and 30 males) were 28 (25–35) years of age. Homozygous sickle cell anemia was by far the most common disease. Baseline between-group differences were noted for core temperature, respiratory rate, and the total pain score; all these differences suggested a greater severity in the group allocated to NIV (Table 1).

Table 1 Baseline characteristics of episodes of acute chest syndrome allocated at random to oxygen or noninvasive ventilation (NIV)
Full size table

The main symptoms at admission are reported in Table 1. Hypoxemia was mild to moderate in most episodes; however, \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) was lower than 65 mmHg in one-third of episodes (12 in the oxygen group and 10 in the NIV group). Hypercapnia, defined as \( {\text{Pa}}_{{{\text{CO}}_{ 2} }} \) greater than 45 mmHg, was noted in nine (25%) episodes in the oxygen group and seven (20%) in the NIV group. Laboratory markers for hemolysis were similar in the two groups. Additional data are given in the Electronic Supplementary Material (Online Resource 2) for the distribution of the \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) values (Fig. E1) and the characteristics of the more hypoxemic patients (Table E1).

Main respiratory changes in the oxygen and noninvasive ventilation groups

The proportion of patients whose \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) was greater than 80 mmHg on day 3 was not significantly different between the two groups [8/32 (25%) with oxygen and 11/31 (35%) with NIV; P = 0.5]. Neither were the \( ({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} \) values different between the two groups on day 3 [24 mmHg (17–33) with oxygen and 21 mmHg (14–26) with NIV; P = 0.22] (Table 2).

Table 2 Changes from baseline to day 3 in 71 episodes of acute chest syndrome (67 patients) allocated at random to oxygen or noninvasive ventilation (NIV)
Full size table

NIV, but not oxygen, was associated with significant decreases in respiratory rate on day 2 [20 (18–24) vs. 24 (20–28); P = 0.006] and day 3 [20 (18–22) vs. 24 (20–28); P = 0.02]. Similarly, significant improvements in \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) and \( ({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} \) from baseline to day 3 were noted with NIV but not with oxygen (Table 2; Fig. 2). Finally, \( {\text{Pa}}_{{{\text{CO}}_{ 2} }} \) remained unchanged over the 3-day period in the NIV group but worsened significantly on day 1 in the oxygen group (Fig. 3). Subgroups analyses are available for the more hypoxemic patients (see Online Resource 3 and Figs. E2, E3, and E4 in the Electronic Supplementary Material).

Fig. 2
figure 2

Changes from baseline in the partial pressure of oxygen in arterial blood \( ( {\text{Pa}}_{{{\text{O}}_{ 2} }} ) \) and the alveolar–arterial oxygen gradient \( (({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} ) \) measured in patients breathing room air in each of the two treatment groups. The medians and the 10th, 25th, 75th, and 90th percentiles are shown as vertical boxes with error bars. Significant improvements from baseline to day 3 were noted in the NIV group for \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) [increase from 69 mmHg (64–75) to 73 mmHg (69–82); P = 0.009] and for \( ({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} \) [decrease from 29 mmHg (23–34) to 21 mmHg (14–26); P = 0.002]. In the oxygen group, no significant changes occurred from baseline to day 3 for \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) [69 mmHg (62–76) and 72 mmHg (62–80), respectively; P = 0.05] or \( ({\text{A}} - {\text{a}})_{{{\text{O}}_{ 2} }} \) [26 mmHg (20–36) and 24 mmHg (17–33), respectively; P = 0.05]

Full size image
Fig. 3
figure 3

Changes in partial pressure of carbon dioxide in arterial blood \( ( {\text{Pa}}_{{{\text{CO}}_{ 2} }} ) \) values in each of the two treatment groups. The medians and the 10th, 25th, 75th, and 90th percentiles are shown as vertical boxes with error bars. Asterisk there was a significant, but not sustained, initial worsening of the \( {\text{Pa}}_{{{\text{CO}}_{ 2} }} \) values in the oxygen group on day 1 (P = 0.004)

Full size image

Invasive mechanical ventilation was not required for any of the ACS episodes included in the study. Changes in vital signs, physical findings, the radiological score, and laboratory variables are shown in Table 2.

Blood transfusions, pain severity, and narcotic use

Red blood cell transfusion requirements were not significantly different in the two groups. The multi-site pain score was higher during the step-down unit stay in the NIV group, whereas the overall pain scores were not different between the two groups. Neither was a difference in the cumulative amounts of narcotics given during the step-down unit stay between the two groups (Table 3).

Table 3 Treatments and outcomes in 71 episodes of acute chest syndrome (67 patients) allocated at random to oxygen or noninvasive ventilation (NIV)
Full size table

Lengths of stay in the step-down unit and hospital

Length of stay in the step-down unit was significantly longer in the NIV group. However, hospital length of stay was not different between the two groups (Table 3). Subgroup analyses for pain relief, narcotic, and blood transfusion requirements, and lengths of stay are available for the more hypoxemic patients (see Online Resource 3 and Table E2 in the Electronic Supplementary Material).

Patient satisfaction and compliance with the respiratory device

The patients’ satisfaction and compliance with the respiratory device are detailed in Online Resource 4 and Fig. E5 in the Electronic Supplementary Material.

Discussion

This prospective, randomized, single-center, open study was designed to evaluate whether NIV improved oxygenation faster than standard care and improved various outcomes, compared to supplemental oxygen alone, in adult patients with early ACS complicating sickle cell disease. NIV provided faster improvements in respiratory rate and gas exchange compared to oxygen alone but failed to significantly reduce the number of patients still hypoxemic at day 3.

ACS is among the leading causes of admission and death in patients with sickle cell disease. The diagnosis of early-stage ACS may be difficult [1, 12] as the clinical presentation consists of nonspecific manifestations such as a fever, chest pain, tachypnea, wheezing, coughing, and/or a new pulmonary infiltrate [2, 4]. Hypoxemia is a feature in about 70% of episodes [13, 14]. In a prospective study of 671 ACS episodes requiring hospital admission in 538 patients, admission was for another reason (mainly pain) in half the episodes, with clinical and radiographic evidence of ACS developing within 3 days after admission [2]. Laboratory values worsened after the diagnosis of ACS despite aggressive treatment, highlighting the need for close monitoring. For our study, we defined early-stage ACS as mild hypoxemia \( ( {\text{Pa}}_{{{\text{O}}_{ 2} }} < 80\,{\text{mmHg)}} \) combined with suggestive clinical and radiographic manifestations or only extrarespiratory symptoms. At diagnosis, the most common clinical manifestations were chest pain, extrathoracic pain, and crackles in both lung fields. A new pulmonary infiltrate was common. Laboratory tests showed moderate-to-severe decreases in hemoglobin and increases in bilirubin and lactic dehydrogenase levels.

Given that the precipitating factor of ACS usually remains unidentified, the optimal treatment for ACS is unclear. However, aggressive treatment is recommended, including hospitalization, analgesics, antibiotics, intravenous fluids, blood transfusions, and supplemental oxygen [2, 4, 15, 16]. Regional alveolar hypoxia and hypoxic vasoconstriction are two major mechanisms responsible for trapping of sickle cells within the small lung vessels [3]. Regional alveolar hypoxia promotes the adhesion of sickle cells to the pulmonary vascular endothelium, whereas regional hypoxic vasoconstriction promotes the mechanical entrapment of rigid sickle cells. Thus, regional alveolar hypoxia may initiate regional occlusion of the pulmonary microvessels. Evidence that correcting the hypoxia reverses sickle cell entrapment [3] suggests that supplemental oxygen should be given to patients with ACS. However, oxygen alone is probably insufficient to improve alveolar oxygenation in regions with marked ventilation–perfusion mismatching [3, 5]. Incentive spirometry alone or combined with intermittent positive end-expiratory pressure (PEEP) or bilevel positive airway pressure was effective in preventing ACS in patients with vaso-occlusive crisis [6–9] and in ensuring resolution of established ACS [10]. Patients with vaso-occlusive crisis and chest pain exhibited rapid shallow breathing with smaller tidal volumes and higher respiratory rates, but similar minute ventilation, compared to patients with pain at other sites [6]. This finding suggested that using NIV to ensure intermittent positive airway pressure might help to reverse ACS by reducing the work of breathing and by increasing the functional residual capacity and lung compliance [17, 18]. In a retrospective study of 9 children with 25 episodes of ACS between 1994 and 2000, bilevel positive airway pressure combined with antibiotics, intravenous fluids, and blood transfusions induced significant improvements in oxygen requirements, heart rate, and respiratory rate [10]. Most of the patients received bilevel positive airway pressure continuously for 72 h then overnight only; mean maximum intermittent positive airway pressure was 12 cmH2O and mean maximum expiratory positive airway pressure 6 cmH2O.

In our study, NIV was used intermittently for 3 days, for a median duration of 4 h/day. Oxygen was set to obtain \( {\text{Sp}}_{{{\text{O}}_{ 2} }} \) value ≥ 95%, as pulse oximetry was found to be more accurate than calculated saturation in the presence of abnormal hemoglobin [19]. A significant decrease in respiratory rate versus baseline was noted as early as day 2 in the NIV group, despite higher baseline values, compared to the oxygen group. In addition, \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) increased versus baseline in the NIV group, suggesting improved gas exchange. Of note, \( {\text{Pa}}_{{{\text{CO}}_{ 2} }} \) remained unchanged over the 3-day period in the NIV group but worsened significantly on day 1 in the oxygen group. Conceivably, NIV induced alveolar recruitment, thereby improving oxygenation.

However, NIV failed to improve the primary outcome measure (number of patients hypoxemic at day 3), and neither did NIV improve pain control, decrease blood transfusion requirements, or decrease length of stay. These results may be related to the limitations of our study. First, our inclusion criteria may have been too broad. Patients were required to have mild-to-moderate hypoxemia with a median \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \) of 69 mmHg (64–75) on room air, and a normal chest X-ray did not prevent inclusion. A more severe hypoxemia, defined with a \( {\text{Pa}}_{{{\text{O}}_{ 2} }} \le 6 5\,{\text{mmHg}} \), was present at baseline in 22 episodes. A more stringent criterion \( ( {\text{Pa}}_{{{\text{O}}_{ 2} }} \le 6 0\,{\text{mmHg)}} \) was present in only 11 episodes (6 in the oxygen group and 5 in the NIV group), a subgroup too small to allow a reliable statistical analysis. Whether NIV is effective in preventing respiratory failure in patients with greater disease severity deserves evaluation. Of note, in the subgroup of patients with severe hypoxemia \( ( {\text{Pa}}_{{{\text{O}}_{ 2} }} \le 6 5\,{\text{mmHg)}} \), physiological variables also improved faster in the NIV group, the differences being slightly more pronounced. Second, baseline differences occurred between the two groups determined by random allocation. These differences indicated a greater severity of ACS in the NIV group (higher core temperature, faster respiratory rate, and higher pain scores), and also, possibly, the longer length of stay. Therefore, they may have led to underestimation of the efficacy of NIV. Next, although the study was conducted in a quaternary referral center that has extensive experience with NIV, the median NIV duration was only 4 h/day, which was perhaps inadequate to ensure efficacy, especially in the more hypoxemic patients with hypercapnia. This might reflect at least in part the difficulty to administer NIV in sickle cell disease patients, who may be reluctant to accept NIV for treating established ACS, as suggested by the least satisfaction and compliance with NIV, as compared with oxygen (see Online Resource 4 and Fig. E5 in the Electronic Supplementary Material). Finally, our study may have been underpowered to show a difference on the main end-point.

In summary, NIV failed to significantly decrease the rate of hypoxemic patients on day 3 in patients with moderately severe early-stage ACS complicating sickle cell disease. Unfortunately, there were baseline differences in pain scores and breathing rates favoring the oxygen group. Moreover, NIV use was only 4 h/day in the NIV group. Despite these limitations, however, there were some benefits of gas exchanges in the NIV group although these did not translate into reduced narcotic and blood transfusion requirements and lengths of stay. Our results should prompt larger study targeting longer periods of NIV in the more severely hypoxemic patients.