Severe Sepsis As Immune Stimulation and Protection from Relapse of Childhood Acute Lymphoblastic Leukemia (ALL): Results from a Pilot Study and Validation of Events in a National Administrative Dataset

Abstract

Relapse remains the major clinical risk in children with ALL, and those with persistent minimal residual disease (MRD) after chemotherapy pose a particular therapeutic challenge given their chemoresistant cell population. We previously showed TLR agonists, an immune-stimulating drug class, generate durable memory immune responses against ALL cells in preclinical MRD models (Blood 2009; 114:2459). These drugs mimic the “danger” signal provided by severe sepsis/septic shock (SS) and may offer immunotherapeutic options for patients who fail or are not eligible for targeted immunotherapies. Demonstrating a protective effect of SS against ALL relapse would further support clinical development of TLR agonists. We hypothesize SS but not SIRS/sepsis (SIRS) will be associated with a decreased risk of ALL relapse, and this effect will be most pronounced in children with NCI high-risk (NCI-HR) disease who are more likely to have both SS exposures and relapses.We assembled and validated a cohort of over 13,000 children with newly diagnosed ALL using the Pediatric Health Information System (PHIS) database. SS exposures after start of induction chemotherapy were defined using ICD9 (NEJM 2003; 348:1546) and billing codes for blood cultures and resources indicating organ failures. ICD9 and pharmaceutical/procedure billing codes were also used to identify ALL relapses and blood/marrow transplantations (BMT). To determine the operating characteristics of these definitions of SS, relapse and BMT, we established a cohort of children from the Children's Hospital of Philadelphia (CHOP) including all patients diagnosed with ALL at CHOP from January 1, 2004 - December 31, 2013. Patients initially treated elsewhere or who left CHOP prior to the end of induction were excluded. Exposures and outcomes identified by manual chart review were considered the gold standard. Exposures were classified as SS or SIRS per the 2005 International Pediatric Sepsis Consensus Guidelines (Ped Crit Care Med 2005; 6:2) and documented until the first event or 3 years from diagnosis. Relapses and deaths were captured until 5 years from diagnosis. Patients were censored at BMT in first remission. We analyzed the data in this single-institution cohort as a pilot study to determine whether the effect of SS on relapse is in the direction we hypothesize. Demographics were evaluated using one-way ANOVA and chi-square. Cox regressions with time-dependent exposures were used to estimate hazard ratios (HR) for relapse by exposure: SS versus no SS, and any sepsis (SIRS + SS) versus none. We also present validation results for the identification of relapses and BMT in the PHIS cohort.Among 395 eligible CHOP patients with newly diagnosed ALL, there were 1736 total exposures (SIRS n=1696 and SS n=40). Demographics are presented in Table 1. SS occurred in 35 patients, SIRS was the maximum exposure in 318, and no sepsis was observed in 42. Exposure was associated with age and phenotype/NCI risk status. Median follow-up was 1826 days (IQR 28-1826), and median time to first SS 117 days (IQR 53-356). There were 521 organisms documented during sepsis episodes (Table 2). BMT in CR1 was the first event in 22 patients (5.6%) at median 174 days from diagnosis (IQR 143-240), relapse in 47 (11.9%; 580 days, 260-1106), and remission death in 5 (1.3%; median 252 days, 98-348). Unadjusted analysis showed a detrimental effect of SS on relapse (HR 2.45; 95%CI 1.04, 5.79). Adjusting for NCI-HR decreased the SS effect size (HR 1.74; 0.7, 4.29), and limiting to NCI-HR patients shifted the effect into a protective direction (crude HR 0.71; 0.09, 5.43). Any sepsis was not protective in any analysis (crude HR 1.86; 0.66, 5.25; adjusted HR 1.71; 0.61, 4.81; NCI-HR-only HR 2.47; 0.32, 19.1). Our relapse identification method within the national PHIS cohort had sensitivity 76% and positive predictive value (PPV) 91%, while BMT identification sensitivity was 93% and PPV 100%.Our CHOP pilot study demonstrated a detrimental effect of SS on the cohort as a whole, suggesting most children with ALL incur more risk than benefit from SS exposure; however, the hazard ratio for SS is in the direction of protection in the NCI-HR group. We have validated our approach to identifying relapse and BMT in PHIS and are validating our method to detect SS exposures in PHIS. Evaluation of the association between SS and relapse risk in the larger, multi-institutional PHIS cohort will provide more precise effect estimates. [Display omitted] DisclosuresFisher:Merck: Research Funding; Pfizer: Research Funding.