The Impact of Steroid Responder Status on Long-Term Outcomes in Critically Ill Patients With Acute Respiratory Distress Syndrome Receiving High-Dose Glucocorticoids

Abstract

Background and objective High-dose intravenous pulsed glucocorticosteroids (GCS) are not part of the standard treatment in acute respiratory distress syndrome (ARDS), and the evidence supporting their use is conflicting. In clinical practice, however, they are used in specialist settings when clinico-patho-radiological features suggest a potentially steroid-responsive pattern, or as a last resort in cases where patients are unable to be weaned off mechanical ventilation. This study aimed to investigate if an early objective response to high-dose GCS treatment in selected critically ill patients is predictive of survival in ARDS. Methods This study involved a case series of 63 patients treated at a tertiary specialist respiratory ICU between 2009 and 2017 who received high-dose GCS for ARDS following a multidisciplinary board agreement. Patients were stratified according to the change in their modified lung injury score (mLIS) between days 0 and 10 following GCS initiation. Changes in mLIS (range: 0-4) were grouped as follows - full responders: ≥2, partial responders: ≥1 and <2, and non-responders: <1. Mortality on discharge and at 6, 12, 18, and 24 months post-ICU discharge was assessed for each group. Data were analysed using logistic regression and a receiver operating curve (ROC) to determine a statistically significant association between the change in mLIS and survival. Results Of the 63 patients, there were seven full responders, 12 partial responders, and 44 non-responders to high-dose GCS. Overall mortality at ICU discharge and 6, 12, 18 and 24 months post-discharge was 29/63 (46.0%), 33/63 (52.4%), 34/63 (54.0%), 34/63 (54.0%), and 35/63 (55.6%) respectively. Mortality was significantly lower in the partial and full-response groups than in the non-response group at all time frames. Logistic regression showed a significant association between the change in mLIS and survival (p<0.001), and a ROC demonstrated that categorising the change in mLIS was a good predictive model for survival (c-statistic 0.86). Conclusions Measuring the change in mLIS by day 10 following high-dose GCS administration for ARDS may be clinically useful in prognosticating such patients. Further research using mLIS as a measure of response to GCS, and larger datasets to enable the evaluation of prognostic factors, may assist clinicians in predicting which patients with persistent ARDS are likely to respond to GCS therapy.

Keywords: acute respiratory distress syndrome [ards]; extracorporeal membrane oxygenation support; glucocorticoid therapy; immunosuppression; invasive mechanical ventilation; lung injury score; mortality predictors; pulmonary critical care; response; systemic steroids.

Figure 1. Consort diagram depicting the inclusion and exclusion of patients

Figure 2. Average mLIS scores: overall and by parameter on days 0 and 10

This figure displays the mean score and 95% confidence intervals for each domain of the Murray LIS as well as the overall mLIS on day 0 and day 10 following high-dose GCS therapy. The scores of each LIS domain are displayed for days 0 and 10. All of these were lower on day 10 than on day 0; A - CXR score (p<0.001, RRb = 1), B - PEEP (0.001, RRb = 0.561), C - P/F ratio (p = 0.008, RRb = 0.775), D - compliance (p<0.001, RRb = 0.638) and E - overall mLIS (p<0.001, RRb = 0.848). Note that the P/F ratio (C) (n = 21) does not include patients on ECMO GCS: glucocorticosteroids; CXR: chest X-ray; P/F ratio: the ratio of arterial oxygen concentration to the fraction of inspired oxygen; RRb: Rank-Biserial Correlations; PEEP: positive end-expiratory pressure; ECMO: extracorporeal membrane oxygenation

Figure 3. Survival following high-dose GCS therapy over time for each patient response group and the overall patient cohort

This graph shows survival at various intervals (ICU discharge and 6, 12, 18, and 24 months post-ICU discharge) for each of the different responder groups as well as the entire cohort (see Appendices) GCS: glucocorticosteroids

Figure 4. The effect of change in mLIS score between day 0 and day 10 on the probability of death

A graph that shows the estimated impact of change in mLIS on the probability of death according to the logistic regression. Real outcomes are displayed by dots for each case, with 0.0 on the Y axis representing survival and 1.0 representing death at 24 months plotted against their measured change in mLIS score

Figure 5. ROC curve for change in mLIS as a predictor for mortality at 24 months

This ROC curve shows the true positive rate against the false positive rate for different thresholds of change in mLIS between day 0 and day 10 as a predictor for survival at 24 months. Predictive tools with potentially high sensitivity and specificity will have high y-axis values for low x-axis values. The c-statistic (area under the curve) calculated from this curve is 0.86 (range: 0-1, with higher values indicating a more reliable predictive tool)