The Association of Corticosteroids and Pediatric Sepsis Biomarker Risk Model (PERSEVERE)-II Biomarker Risk Stratification With Mortality in Pediatric Septic Shock* : Pediatric Critical Care Medicine

Journal Logo

Feature Articles

The Association of Corticosteroids and Pediatric Sepsis Biomarker Risk Model (PERSEVERE)-II Biomarker Risk Stratification With Mortality in Pediatric Septic Shock*

Klowak, Jennifer A. MD, MSc1; Bijelić, Vid MSc2; Barrowman, Nick PhD2; Menon, Kusum MD, MSc1,2;  for the Genomics of Pediatric Septic Shock Investigators

Author Information
Pediatric Critical Care Medicine 24(3):p 186-193, March 2023. | DOI: 10.1097/PCC.0000000000003117



  • Pediatric sepsis remains a common cause of pediatric critical care admission and represents a global health burden.
  • Given the lack of effective targeted treatments despite extensive research, there is increasing interest in prognostic enrichment for treatments such as corticosteroids.
  • Corticosteroids may be associated with harm in a subgroup of children with a high baseline risk of mortality, assessed using the PERSEVERE-II risk score.


  • Studies on corticosteroids in pediatric sepsis have yielded mixed results. An understanding of which subgroups may benefit or be harmed by corticosteroids could inform decision-making.
  • In this multicenter observational study, corticosteroid administration was associated with an increased risk of mortality in a subgroup of children with a high baseline PERSEVERE-II risk score.
  • The PERSEVERE-II risk score has the potential to inform prognostic enrichment; however, future prospective and randomized trials in pediatric sepsis are needed to confirm our findings.

Sepsis in children remains a leading global health challenge (1). Septic shock accounts for approximately 8% of PICU admissions (1), carries a high mortality rate of 15–25% (1–3), and leaves a third of survivors with persistent disability. Despite rigorous research, there are no targeted treatments for pediatric sepsis, and management of septic shock currently consists of antibiotic, fluid, and vasopressor administration (4). Given the ongoing mortality and morbidity, clinicians look to adjunctive therapies such as corticosteroids to improve outcomes.

Corticosteroids are widely used to treat pediatric septic shock despite inadequate evidence to support this practice (5–7). A limited number of small studies suggest a shorter time to shock reversal with corticosteroid use. However, other, mostly observational studies found an association of corticosteroids with increased mortality (8), increased risk of secondary infections (9), and hyperglycemia, hypernatremia, and muscle wasting (10). These contradictory results may reflect numerous considerations including differing study populations, variable dose and duration of corticosteroids (7), and individual differences in corticosteroid gene expression–based risk stratification (11).

Mortality risk stratification may identify a subset of patients who benefit from or are harmed by corticosteroid administration. The Pediatric Sepsis Biomarker Risk Model (PERSEVERE) score is a biomarker-based mortality risk stratification tool for pediatric sepsis, with its roots in discovery-oriented transcriptomic studies (12). A retrospective analysis of 496 children with septic shock adjusting for the original PERSEVERE score did not find a relationship between corticosteroids and mortality (13). PERSEVERE-II score is based on five sepsis-related biomarkers (C-C chemokine ligand 3, interleukin-8, heat shock protein 70 kDa 1B, granzyme B, and matrix metallopeptidase 8) as well as platelet count (14). In a recent validation study, PERSEVERE-II showed improved performance (area under the curve 0.83) over PERSEVERE-I at discriminating 28-day mortality (15).

Therefore, our objective was to explore whether corticosteroid administration within PERSEVERE-II risk categories was associated with 28-day mortality, ICU-free days, and maximum number of organ failures in children with septic shock.


We conducted a secondary analysis of data obtained from a previous prospective multicenter cohort study of 461 children with septic shock (15).


Children admitted to the PICU who met pediatric-specific consensus criteria for septic shock (16) upon admission. Children were excluded if the care team felt removal of an additional 7.5 mL of blood was contraindicated, there was no stable central venous or arterial catheter access or plans for routine phlebotomy. Enrollment occurred between January 2015 and December 2018 and included participants from 13 centers across the United States.

Data Collection

Clinical and demographic data from patient charts were recorded at study entry, and patients were followed for up to 28 days or death. Corticosteroid administration was defined as receipt of any formulation of systemic corticosteroids for at least 48 hours during the initial 7 days of septic shock and excluding steroids administered only periextubation. Blood samples were drawn on the first day of enrollment (designated day 0) with clinical samples. The research blood samples underwent analysis at the primary study site, Cincinnati Children’s Hospital Medical Center, for the components of the PERSEVERE-II score as previously described (14,15). The baseline PERSEVERE-II score was used to classify patients into the mortality risk strata: low risk (baseline mortality probability estimated at < 2%) and high risk (baseline mortality probability estimated at 16.7–57.1%). The scoring system rules were locked prior to the study start. Outcome assessors and clinicians were not aware of the PERSEVERE-II score. Complicated course is an empirically derived composite outcome used in previous work (17,18) and defined as death by 28 days or at least two organ failures (pediatric consensus criteria) (16) persisting on day 7 of septic shock.


Institutional Review Board (IRB) approval was attained from all trial participating centers (Cincinnati Children’s Hospital IRB 2008-0558), and informed consent including for future data analyses was obtained prior to enrollment in the original prospective study. We attained local IRBs approval (CHEO IRB 21/75X) for secondary data analysis.

Data Analysis

Descriptive data were summarized using frequencies and percentages or medians (interquartile ranges [IQRs]) as appropriate. All point estimates are reported with 95% CIs. A two-sided p value of less than 0.05 was considered statistically significant. All analyses were performed using R software Version 4.0.5 (R Core Team, 2021, Vienna, Austria).

Our primary objective was to assess the relationship between corticosteroid administration and mortality across PERSEVERE-II baseline mortality risks. We performed a logistic regression for 28-day mortality with six a priori covariates: corticosteroid administration, PERSEVERE-II risk category, age (< 1 or over 1 yr), malignancy, comorbidity, and Pediatric Risk of Mortality Score (PRISM III) (19), as well as an interaction between corticosteroids and PERSEVERE-II risk. We calculated odds ratios (ORs) and marginal mean probability with 95% CIs. Given significant interaction between corticosteroids and PERSEVERE-II score, ORs were reported by PERSEVERE-II risk subcategory. Analysis was performed with and without multiple imputation by chained equations using predictive mean matching (20), and results are reported using the imputed dataset. A subgroup analysis excluding patients with bone marrow transplantation and malignancy was performed for the primary outcome of 28-day mortality using Firth’s logistic regression due to limited sample size.

As an alternative approach to address potential confounding, we calculated propensity scores based on centre, age, malignancy, comorbidity and PRISM score and used inverse probability treatment weighting (IPTW) to estimate the average treatment effect on the treated (ATT) (21).

For the secondary outcome ICU-free days, a weighted survival analysis was performed with the IPTW population. Logistic regression was used for complicated course, as above. The secondary outcome, maximum failed organs, was log transformed due to skewed distribution prior to linear regression.


Clinical Characteristics

Table 1 describes the characteristics of patients who received (n = 215; 46.6%) and did not receive corticosteroids (n = 246; 53.4%). Corticosteroid recipients were more likely to have comorbidities (76.3% vs 65.4%), malignancy (23.3% vs 6.9%), immunosuppression (40.0% vs 24.5%), higher PERSEVERE-II score (47.0% vs 30.1%), and higher PRISM score (median 12 [IQR 8–18] vs 9 [5–13]). Missing data were less than 1% except for PRISM score (8%).

TABLE 1. - Patient Demographics and Clinical Characteristics
Variable Overall, N = 461 Corticosteroids, N = 215 No Corticosteroids, N = 246
Age (yr), median (IQR) 7.1 (2.2–13.6) 7.3 (2.0–13.3) 7.0 (2.4–14.2)
Age (yr), n (%)
 < 1 68 (14.8) 28 (13.0) 40 (16.3)
 ≥ 1 393 (85.2) 187 (87.0) 206 (83.7)
Pediatric Sepsis Biomarker Risk Model-II score
 Low risk 286 (62.0%) 114 (53.0) 172 (69.9)
 High risk 175 (38.0%) 101 (47.0) 74 (30.1)
Pediatric Risk of Mortality III score, median (IQR) 11 (6–16) a 12 (8–18) a 9 (5–13) a
Comorbidity presence, n (%) 325 (70.5) 164 (76.3) 161 (65.4)
Malignancy presence, n (%) 67 (14.5) 50 (23.3) 17 (6.9)
Immunosuppression, n (%) 113 (24.5) 86 (40.0) 27 (11.0)
IQR = interquartile range.
an = 424 for Pediatric Risk of Mortality II (overall), n = 204 (corticosteroids), n = 220 (no corticosteroids).

Corticosteroids and Mortality

Table 2 shows the logistic regression for the outcome of 28-day mortality. There were 58 patients who died (58/461; 12.6%). Imputation for missing data did not have a significant effect on the results. In the overall cohort, the association of corticosteroid exposure and mortality varied by PERSEVERE-II risk status (p = 0.01). Children with a baseline low mortality risk by PERSEVERE-II did not have an increased mortality associated with corticosteroid exposure (OR 0.20 [95% CI 0.02–1.73]; p = 0.15; n = 286); however patients, with a high-risk PERSEVERE-II score had an increased odds of mortality with corticosteroid exposure (OR 4.10 [95% CI 1.70–9.86]; p = 0.002; n = 175) (Table 2). To assess the robustness of our finding across analytical methods, IPTW logistic regression was conducted (eTable 1,, which showed an increased odds of mortality with corticosteroid exposure in the high-risk PERSEVERE-II group (OR 2.93 [95% CI 1.14–7.49]; p = 0.03).

TABLE 2. - Multivariate Relationship Between Potential Risk Factors and Mortality in Children With Septic Shock
Variables OR (95% CI) p
Corticosteroid exposure
 With low PERSEVERE-II score 0.20 (0.02–1.73) 0.15
 With high PERSEVERE-II score 4.10 (1.70–9.86) 0.002
High PERSEVERE-II score 2.60 (0.86–7.91) 0.09
Age over 1 yr 0.50 (0.20–1.24) 0.13
Pediatric Risk of Mortality III score 1.08 (1.04–1.13) < 0.001
Comorbidity presence 1.33 (0.58–3.09) 0.50
Malignancy presence 0.77 (0.31–1.88) 0.57
OR = odds ratio, PERSEVERE = Pediatric Sepsis Biomarker Risk Model.

Patients Without Malignancy or Bone Marrow Transplant

Among children without malignancy or bone marrow transplant, the association between corticosteroid and mortality varied by PERSEVERE-II risk status (p = 0.01). Those with a high PERSEVERE-II risk who received corticosteroids had a significantly higher odds of dying than those with a low risk who also received corticosteroids (OR 2.67 [1.03–6.94] vs OR 0.10 [0.00–0.92]) (eTable 2, Corticosteroids did not affect survival in the overall subgroup without corticosteroid-PERSEVERE-II interaction (OR 1.35 [95% CI, 0.60–3.05]; p = 0.47).

Corticosteroids and ICU-Free Days

Corticosteroid exposure in patients with high PERSEVERE-II risk was associated with fewer ICU-free days (p < 0.0001) (Fig. 1).

Figure 1.:
Weighted survival analysis for ICU-free days by corticosteroid exposure and Pediatric Sepsis Biomarker Risk Model-II score (PIIScore).

Corticosteroids and Complicated Course

The association of corticosteroids and complicated course varied by PERSEVERE-II score (p = 0.01). Patients with a high PERSEVERE-II risk who received corticosteroids had a significantly higher odds of a complicated course than those with a low PERSEVERE-II risk who also received steroids (OR 7.02 [95% CI 3.16–15.59]; p < 0.001 vs OR 1.84 [95% CI, 0.91–3.74]; p = 0.09) (Table 3 and Fig. 2).

TABLE 3. - Relationship Between Potential Risk Factors and Development of a Complicated Course or Maximum Number of Failed Organs in Children With Septic Shock
Variables Complicated Course Maximum Number of Failed Organs
OR (95% CI) p OR (95% CI) p
Corticosteroid exposure
 With low PERSEVERE-II score 1.84 (0.91–3.74) 0.09 1.06 (0.95–1.17) 0.29
 With high PERSEVERE-II score 7.02 (3.16–15.59) < 0.001 1.46 (1.26–1.66) < 0.001
High PERSEVERE-II score 1.37 (0.62–3.06) 0.44 1.09 (0.96–1.23) 0.17
Age over 1 yr 0.31 (0.16–0.62) < 0.001 0.81 (0.72–0.92) < 0.001
Pediatric Risk of Mortality III score 1.10 (1.06–1.15) < 0.001 1.03 (1.02–1.03) < 0.001
Comorbidity presence 1.40 (0.76–2.57) 0.28 1.02 (0.93–1.13) 0.64
Malignancy presence 0.49 (0.22–1.07) 0.07 0.92 (0.81–1.04) 0.18
OR = odds ratio, PERSEVERE = Pediatric Sepsis Biomarker Risk Model.

Figure 2.:
Estimated marginal mean probability by corticosteroid exposure and Pediatric Sepsis Biomarker Risk Model (PERSEVERE)-11 risk of 28-d mortality (A) and a complicated course (B). PIIScore = PERSEVERE-11 score.

Corticosteroids and Maximum Number of Failed Organs

The association of corticosteroids and number of failed organs varied by PERSEVERE-II score (p < 0.001). Children with a high PERSEVERE-II score who received corticosteroids were more likely to have a higher number of failed organs (OR 1.46 [95% CI 1.26–1.66]; p < 0.001) than those with a low PERSEVERE-II score who also received corticosteroids (OR 1.06 [95% CI 0.95–1.17]; p < 0.29) (Table 3).


Our study explored the association between corticosteroids and outcomes in a multicenter, heterogeneous cohort of 461 children with septic shock. Importantly, corticosteroid exposure in the overall cohort did not have a significant association with mortality. However, after stratifying for baseline risk of mortality with the PERSEVERE-II score, the children with a high baseline risk who were administered corticosteroids had a higher odds of death (OR 4.10 [95% CI 1.70–9.86]; p = 0.002), complicated course (OR 7.02 [95% CI 3.16–15.59]; p < 0.001), more failed organs (OR 1.46 [95% CI 1.26–1.66]; p < 0.001), and less ICU-free days (p < 0.0001). This association of increased mortality with corticosteroid exposure was consistent across all outcomes as well as multiple statistical analysis techniques, suggesting a potential differential response to corticosteroid administration in sub-groups of children with septic shock.

The difference in the association with mortality of the overall cohort and those in the high PERSEVERE-II risk group supports the premise that children at baseline might be at variable risk of mortality from septic shock as well as responds differently to treatments such as corticosteroids. However, the intrinsic factors that contribute to this variability are not well understood. Previous literature has focused primarily on the association of illness severity scores and mortality with septic shock (22–25). However, these scores may not be sensitive enough to predict those at risk of a differential outcome following therapies such as corticosteroids leading to the impetus for patient specific biomarkers. Our results suggest that PERSEVERE-II biomarkers may predict which septic shock patients are at increased risk of a poor outcome from corticosteroids. PERSEVERE-II biomarkers include gene expression patterns of five proteins related to inflammation and host response to sepsis, but the underlying mechanism by which risk stratification strata relate to outcomes or corticosteroid response is still under investigation (15). It is interesting to note, that even in the subgroup excluding oncology, those in the high PERSEVERE-II risk stratification group had a significantly increased risk of mortality with corticosteroids again reflecting the predictive enrichment capabilities of PERSEVERE-II with respect to corticosteroids.

Our previous retrospective analysis of 496 children with septic shock did not find a relationship between corticosteroids and mortality when stratifying by PERSEVERE-I score (13). The PERSEVERE-II score, however, has improved prognostic accuracy (15) and differs from PERSEVERE-I by inclusion of the platelet count. A recent large meta-analysis found that in septic patients, platelet counts were significantly lower in nonsurvivors versus survivors (26), thus lending support to the premise that inclusion of the platelet count in PERSEVERE-II improved its prognostic accuracy. Enrichment using endotypes, biological subclasses of sepsis, has also been shown to outline a group of children (endotype A) where corticosteroids are associated with mortality in sepsis (27). Our group has previously combined predictive and prognostic enrichment using both PERSEVERE-I and endotype data to identify a group of children in which corticosteroids were associated with more than a 10-fold reduction in the risk of complicated course (17). The combined enrichment capabilities of PERSEVERE-II (prognostic) and endotypes (predictive) were not explored in this study due to insufficient numbers of patients with measured endotypes and remain an area for future study. Prospective validation of our findings may serve to inform future therapeutic randomized controlled trials in pediatric sepsis.

This study has several limitations. The administration of corticosteroids was not standardized, and therefore, it is possible that those in the high risk PESEVERE-II strata received differential corticosteroid dosing. However, given that the study was conducted in 13 different centers all of whom were unaware of the PERSEVERE-II status of their patients, it is unlikely that there was a systematic bias toward higher doses of corticosteroids in the high risk group. This is an observational study, and therefore, residual confounding cannot be excluded.


In our observational multicenter study, we identified an association between corticosteroid administration and increased mortality in pediatric sepsis through prognostic subgrouping using the PERSEVERE-II score. Overall, our study did not find evidence to support the use of corticosteroids in pediatric septic shock. Our findings require prospective validation and may inform future therapeutic randomized controlled trials.


We would like to acknowledge the tremendous work and research legacy of Dr. Hector Wong, who passed shortly prior to article preparation.


1. Weiss SL, Fitzgerald JC, Pappachan J, et al.: Global epidemiology of pediatric severe sepsis: The sepsis prevalence, outcomes, and therapies study. Am J Respir Crit Care Med. 2015; 191:1147–1157
2. Zimmerman JJ, Williams MD: Adjunctive corticosteroid therapy in pediatric severe sepsis: Observations from the RESOLVE study. Pediatr Crit Care Med. 2011; 12:2–8
3. Balamuth F, Weiss SL, Neuman MI, et al.: Pediatric severe sepsis in U.S. children’s hospitals. Pediatr Crit Care Med. 2014; 15:798–805
4. Weiss SL, Peters MJ, Alhazzani W, et al.: Surviving sepsis campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. Pediatr Crit Care Med. 2020; 21:E52–106
5. Carmean A, Fortenberry JD, McCracken C, et al.: A survey of attitudes and practices regarding the use of steroid supplementation in pediatric sepsis. Pediatr Emerg Care. 2015; 31:694–698
6. Menon K, McNally JD, Choong K, et al.: A survey of stated physician practices and beliefs on the use of steroids in pediatric fluid and/or vasoactive infusion-dependent shock. Pediatr Crit Care Med. 2013; 14:462–466
7. Menon K, McNally D, Choong K, et al.: A systematic review and meta-analysis on the effect of steroids in pediatric shock. Pediatr Crit Care Med. 2013; 14:474–480
8. Markovitz BP, Goodman DM, Watson RS, et al.: A retrospective cohort study of prognostic factors associated with outcome in pediatric severe sepsis: What is the role of steroids? Pediatr Crit Care Med. 2005; 6:270–274
9. Menon K, Dayre McNally J, Choong K, et al.: A cohort study of pediatric shock: Frequency of corticosteriod use and association with clinical outcomes. Shock. 2015; 44:402–409
10. Annane D, Bellissant E, Bollaert PE, et al.: Corticosteroids for treating sepsis in children and adults. Cochrane Database Syst Rev. 2019; 12:CD002243
11. Wong HR, Cvijanovich NZ, Anas N, et al.: Developing a clinically feasible personalized medicine approach to pediatric septic shock. Am J Respir Crit Care Med. 2015; 191:309–315
12. Wong HR, Salisbury S, Xiao Q, et al.: The pediatric sepsis biomarker risk model. Crit Care. 2012; 16:R174
13. Atkinson SJ, Cvijanovich NZ, Thomas NJ, et al.: Corticosteroids and pediatric septic shock outcomes: A risk stratified analysis. PLoS One. 2014; 9:e1127027–e1127013
14. Wong HR, Cvijanovich NZ, Anas N, et al.: Pediatric sepsis biomarker risk model-II: Redefining the pediatric sepsis biomarker risk model with septic shock phenotype. Crit Care Med. 2016; 44:2010–2017
15. Wong HR, Caldwell JT, Cvijanovich NZ, et al.: Prospective clinical testing and experimental validation of the pediatric sepsis biomarker risk model. Sci Transl Med. 2019; 11:eaax9000
16. Goldstein B, Giroir B, Randolph A: International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med. 2005; 6:2–8
17. Wong HR, Atkinson SJ, Cvijanovich NZ, et al.: Combining prognostic and predictive enrichment strategies to identify children with septic shock responsive to corticosteroids. Crit Care Med. 2016; 44:e1000–e1003
18. Banerjee S, Mohammed A, Wong HR, et al.: Machine learning identifies complicated sepsis course and subsequent mortality based on 20 genes in peripheral blood immune cells at 24 H Post-ICU admission. Front Immunol. 2021; 12:592303
19. Pollack MM, Patel KM, Ruttimann UE: PRISM III: An updated pediatric risk of mortality score. Crit Care Med. 1996; 24:743–752
20. van Buuren S: Flexible Imputation of Missing Data. Second Edition. Boca Raton, FL, Taylor and Francis Group, 2018
21. Elze MC, Gregson J, Baber U, et al.: Comparison of propensity score methods and covariate adjustment: Evaluation in 4 cardiovascular studies. J Am Coll Cardiol. 2017; 69:345–357
22. Khan MR, Maheshwari PK, Masood K, et al.: Epidemiology and outcome of sepsis in a tertiary care PICU of Pakistan. Indian J Pediatr. 2012; 79:1454–1458
23. Goh AYT, Chan PWK, Lum LCS: Sepsis, severe sepsis and septic shock in paediatric multiple organ dysfunction syndrome. J Paediatr Child Health. 1999; 35:488–492
24. Breuling T, Tschiedel E, Groe-Lordemann A, et al.: Septic shock in children in an urban area in Western Germany--Outcome, risk factors for mortality and infection epidemiology. Klin Padiatr. 2015; 227:61–65
25. Ibrahiem SK, Galal YS, Youssef MRL, et al.: Prognostic markers among Egyptian children with sepsis in the intensive care units, Cairo University Hospitals. Allergol Immunopathol (Madr). 2016; 44:46–53
26. Menon K, Schlapbach LJ, Akech S, et al.: Criteria for pediatric sepsis-A systematic review and meta-analysis by the pediatric sepsis definition taskforce. Crit Care Med. 2022; 50:21–36
27. Wong HR, Cvijanovich NZ, Anas N, et al.: Endotype transitions during the acute phase of pediatric septic shock reflect changing risk and treatment response. Crit Care Med. 2018; 46:e242–e249

biomarkers; children; corticosteroids; mortality; sepsis; septic shock

Supplemental Digital Content

Copyright © 2023 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies