Can We Predict Acute Respiratory Distress Syndrome in Hematopoietic Stem Cell Recipients?

Immunocompromised patients represent up to one-third of critically ill patients and one-fifth of those admitted to the ICU with acute respiratory distress syndrome (ARDS) (1). This group has distinct risk factors for lung injury and increased mortality compared with nonimmunocompromised patients with ARDS, demanding expertise for clinical management. Among critically ill immunocompromised patients, hematopoietic stem cell transplantation (HCT) recipients exhibit the highest mortality (2). Three significant differences set HCT recipients apart from other immunocompromised patients. First, factors like transplant type, conditioning regimen, and underlying malignancy contribute to clinical heterogeneity (3). Second, allogeneic HCT recipients may manifest unique complications, including graft-versus-host disease (GVHD), sinusoidal obstruction syndrome, acute interstitial pneumonia, diffuse alveolar hemorrhage, and bronchiolitis obliterans, among others (4). Third, varying degrees of immunosuppression can lead to opportunistic infections, making ARDS determinants in this group distinctive (Figure 1). Despite general progress, HCT recipients remain a subgroup requiring focused attention for ARDS prognosis and management.

Figure 1.

Determinants of acute respiratory distress and lung injury among hematopoietic stem cell transplantation recipients. The figure highlights key determinants of acute respiratory distress syndrome and acute lung injury in hematopoietic stem cell transplantation recipients. Examining these factors is crucial when considering the differential diagnosis of acute respiratory failure in this population. Acute respiratory failure in these patients has distinctive infectious and noninfectious etiologies (e.g., periengraftment respiratory distress syndrome, idiopathic pneumonia syndrome, diffuse alveolar hemorrhage, graft-vs.-host disease, drug toxicity, etc.). Figure was created with BioRender.com and partially inspired by a previous figure available at https://clinicalgate.com/pulmonary-complications-of-hematopoietic-stem-cell-transplantation/.

ICU admission occurs in approximately 20% of HCT recipients, with a higher incidence among allogeneic recipients (5–7). Acute respiratory failure is the predominant diagnosis at ICU admission, with more than half of these patients requiring invasive mechanical ventilation (5, 7). Historically, hospital mortality in HCT recipients receiving invasive mechanical ventilation was reported as high as 90%, questioning its futility (8). Although critically ill patients with hematologic malignancies have shown improved survival (9, 10), this trend is not consistently observed in HCT recipients (2, 11). Currently, non–relapse-related mortality in HCT recipients is mainly secondary to pulmonary complications and acute respiratory failure (12). Given these considerations, a pragmatic approach focuses on prevention and early interventions. Given the devastating consequences of ARDS in this population, clinicians grapple with key questions: Can we accurately identify HCT recipients who are at high risk of developing ARDS? And, crucially, what interventions are warranted for those identified as being at highest risk?

In this issue of the Journal, Herasevich and colleagues (pp. 543–552) present the results of a multicenter cohort study describing the derivation and validation of an ARDS prediction model in adult HCT recipients: the Lung Injury Prevention Score for Bone Marrow Transplant Patients (LIPS-BMT) (13). The study population included autologous and allogeneic HCT recipients who required hospitalization from three different Mayo Clinic sites. One site served as the derivation cohort (888 patients, 1,718 hospitalizations), and the other two served as the validation cohort (470 patients, 1,005 hospitalizations). The primary outcome was the development of ARDS. Secondary outcomes were the receipt of mechanical ventilation and acute respiratory failure. Predictors were categorized into pretransplant, post-transplant, and in-hospital domains. ARDS incidence was 1.7% and 3.5% in the derivation and the validation cohorts, respectively. The incidence of ICU admission was 15% in both cohorts. The final model consisted of a 22-point score with very good accuracy for predicting ARDS in the validation cohort (area under the receiver operating characteristic curve, 0.85; 95% confidence interval [CI], 0.79–0.90). The predictive accuracy was 0.87 (95% CI, 0.83–0.91) and 0.71 (95% CI, 0.68–0.75) for the receipt of mechanical ventilation and acute respiratory failure, respectively.

Authors should be commended for investigating this relevant clinical question. Moreover, the study design, sample size, and inclusion in the model of variables that depict distinct periods within the transplantation procedure are strengths of this study. The resulting LIPS-BMT score is performant and enables us to discriminate almost impeccably among those with and without ARDS risk. The reproducibility of the methods used instills confidence in the study findings, further bolstered by the online publication of its code. These practices should be highly encouraged. Furthermore, the participating centers have an extensive clinical expertise in the field, with a high volume of HCT recipients providing precise estimates of ARDS incidence.

However, the LIPS-BMT score exhibits certain limitations. First, when diagnosing ARDS in HCT recipients, it is important to note that some patients may lack the typical feature of diffuse alveolar damage associated with ARDS. Allogeneic HCT recipients may also experience pulmonary involvement related to GVHD and non-GVHD causes. Assuming these noninfectious events are ARDS or can be predicted by the same models may be misleading. Second, age and sex, which are traditionally associated with lung infection, are missing in the model (6). Third, despite the large number of screened patients, ARDS incidence was low, with only 30 cases observed out of 1,718 hospitalizations. This scarcity of events raises concerns about the score’s generalizability. Fourth, although the LIPS-BMT excels in predicting “ARDS” it offers limited insights into the underlying etiology, hindering the identification of specific preventive or targeted therapeutic strategies. Identifying the underlying cause of lung involvement is crucial for tailoring interventions in this setting (10). Last, specifying GVHD solely as a binary variable might be oversimplifying. The time between GVHD onset and ICU admission and corticosteroid responsiveness are crucial factors influencing outcome (6, 14, 15).

The potential clinical implications of the LIPS-BMT score are twofold: first, the identification of patients at higher risk for deterioration can inform preemptive admission to high-dependency units or ICUs for close monitoring. Second, the tool has the potential to identify patients suitable for preventive therapeutic strategies for ARDS. These two purposes still warrant further discussion. By nature, many of these patients are hospitalized in the hematology or general wards at the time complications occur. In this setting, the score could help stratify the risk of developing further respiratory deterioration, mandating either a closer follow-up, rapid response team involvement, or admission to a more highly monitored setting. A cutoff point with a higher specificity could be preferable for this setting. As such, this tool might be used primarily by hematologists and hospitalists, prompting important discussions with intensivists when alarms are raised according to local thresholds for ICU admission. The potential role of using the LIPS-BMT score for identifying patients suitable for preventive strategies is indeed attractive. However, given the lack of insight in pinpointing the underlying ARDS etiology, the use of LIPS-BMT for this purpose might remain limited. A more nuanced exploration of ARDS etiology—distinguishing between infectious and immunologic causes—is essential to shape these strategies effectively in the future.

In conclusion, this study offers a new perspective on ARDS prediction in HCT recipients. Applying this score to diverse HCT cohorts will refine its effectiveness and aid in developing prevention strategies for a disease still associated with high mortality.