Your browser doesn't support javascript.
Novel EHR-Derived SARS-CoV-2 Pneumonia Subphenotypes Show Divergent Clinical Outcomes with Differential Responses to Corticosteroids
American Journal of Respiratory and Critical Care Medicine ; 205(1), 2022.
Article in English | EMBASE | ID: covidwho-1927886
ABSTRACT

Introduction:

In COVID-19-related acute respiratory distress syndrome (ARDS), two distinct subphenotypes have been identified with differential outcomes and responses to corticosteroid therapy. We aimed to evaluate (1) whether clinical data can identify subgroups in a broader group of patients with SARS-CoV-2 pneumonia and (2) the extent to which corticosteroids demonstrate heterogeneity of treatment effect across such subgroups.

Methods:

We retrospectively studied all SARS-CoV-2 patients hospitalized for >24 hours and requiring oxygen support across 11 BJC HealthCare hospitals from June-December 2020. We excluded the initial surge (March-May 2020), as clinical care was heterogeneous and corticosteroid use low during this period. Using prespecified routinely-collected vital sign and laboratory indicator variables, we sought distinct clinical subphenotypes of SARS-CoV-2 pneumonia through latent class analysis (LCA). Across LCA subphenotypes, we evaluated the relationship between corticosteroid treatment and patient outcomes. We used multivariable logistic regression (dependent variable = composite of death/hospice) to explore treatment interaction between corticosteroid exposure and LCA subphenotype, adjusting for age and maximal SOFA score within 24 hours of admission as surrogates for indication.

Results:

The 3-class LCA model best fit the 1845-patient cohort (p=0.007). Class-1 (n=1456) had mean standardized values of all indicator variables near zero;Class-2 (n=235) manifested profound isolated hypoxemia;and Class-3 (n=154) displayed multiorgan failure, shock, and neutrophilia (Figure-1A). Despite representing <25% of the cohort, Classes 2 (n=109, 46%) and 3 (n=70, 46%) comprised >50% of the primary outcome (vs Class-1 n=151, 10%;p<0.001). Corticosteroids were more frequently administered in Class-2 (n=215, 91%) than in Class-1 (n=1071, 74%) or Class-3 (n=110, 71%, p<0.001;Figure-1B). Adjusted analyses demonstrated interaction between LCA class and corticosteroid treatment for the primary outcome (Class-1, p=0.003;Class-3, p=0.002). Corticosteroids were associated with increased adjusted odds for the primary outcome in Class-1 (aOR 2.11, 95% CI 1.33-3.50, p=0.002) and decreased adjusted odds for the primary outcome in Class-3 (aOR 0.44, 95% CI 0.19-0.98, p=0.048). Class-2 showed no outcome differences between the corticosteroid and noncorticosteroid groups (aOR 1.03, 95% CI 0.38-2.81, p=0.95).

Conclusions:

Three distinct subphenotypes of SARS-CoV-2 pneumonia demonstrate different clinical outcomes and corticosteroid response. No clear effect was seen among patients with isolated hypoxemia, whereas those with multiorgan failure appeared to benefit. Our findings suggest that among hospitalised patients in our healthcare system, corticosteroid therapy was associated with increased risk of harm. Prospective studies are needed to evaluate the efficacy of corticosteroids in SARS-CoV-2 pneumonia in predictively-enriched trials.
Keywords

Full text: Available Collection: Databases of international organizations Database: EMBASE Type of study: Prognostic study Language: English Journal: American Journal of Respiratory and Critical Care Medicine Year: 2022 Document Type: Article

Similar

MEDLINE

...
LILACS

LIS


Full text: Available Collection: Databases of international organizations Database: EMBASE Type of study: Prognostic study Language: English Journal: American Journal of Respiratory and Critical Care Medicine Year: 2022 Document Type: Article