Related Topics

    Model developed to measure risk posed by fever, low white cell counts

    Aditya Gaur, PhD

    A model developed by St. Jude researchers is showing early promise as a tool to help physicians identify cancer patients with low white cell counts who are at greatest risk for complications when feverish.

    The goal is to improve the ability of physicians to distinguish between patients who need hospitalization or treatment with intravenous (IV) antibiotics from those who can safely be treated as outpatients or with antibiotic pills. Such models are already used to help guide treatment of adult cancer patients.

    Currently all children with cancer who develop fever with neutropenia are hospitalized and treated with IV antibiotics. Neutropenia occurs when levels of infection-fighting white blood cells known as neutrophils fall to dangerously low levels.

    “But we know not every patient with fever and neutropenia is at the same risk for developing infection and complications,” said Aditya Gaur, MD, Infectious Diseases, and the paper’s senior author. “If there were a way to identify those most likely to develop complications, perhaps the other patients could undergo an easier therapy on an outpatient basis.”

    Hana Hakim, MD, a research associate in Infectious Diseases, said being treated outside the hospital would be easier on patients and their families. Outpatient care would also help control costs and protect patients from hospital-acquired infections, she said.

    The research was published in the December 4 online edition of the Pediatric Infectious Disease Journal.

    The study is the first of three planned trials to develop, validate and test a tool to help doctors match treatment of fever and neutropenia with the likelihood that a young cancer patient will go on to develop widespread infection or serious clinical complications, including kidney failure or heart and breathing problems. The second phase is already underway at St. Jude and is expected to last another year. Depending on those and future results, a screening test might be available for widespread use in about five years, investigators said.

    “The key factor in risk prediction is to consistently and reliably identify patients at risk,” the investigators wrote. “Of the several risk prediction models proposed, few apply to pediatric patients.”

    For this study, investigators reviewed the medical records of 332 St. Jude cancer patients hospitalized for fever and neutropenia during a two-year period. The scientists used a variety of statistical tools to analyze 15 risk factors. Four emerged as independent predictors of patients who developed clinical complications or widespread infection diagnosed as either invasive bacterial infection or culture-negative sepsis. Additional analysis determined the importance of each of the four risk factors; researchers assigned each a numerical value and used the sum to rank patients as being at low or high risk.

    The factors identified as independently linked to a greater risk for widespread infection were a diagnosis of acute myeloid leukemia (AML), a neutrophil count of less than 100 cells per micro liter, an initial temperature of 102.2 degrees or greater and a sick appearance.

    AML and looking sick were also associated with a greater risk of developing clinical complications. Relapse and race were also predictors. Patients whose cancer had returned and non-white patients were both at higher risk for clinical complications.

    The developed risk prediction tool correctly classified approximately 97 percent of patients who proved to be at low risk for clinical complications and about 96 percent of those at low risk for widespread infection. The method’s predictive value for identifying low-risk patients rose even higher when patients were re-evaluated after 24 hours. In that setting, only one of 223 patients was incorrectly classified as at low risk for widespread infection. The method also mistakenly classified only six out of 207 patients as being at low risk for clinical complications when in fact they were at high risk.

    The findings suggest that patients with fever and neutropenia might benefit from a 24-hour observation period at the start of treatment.

    Hakim said this study is the first that taps clinical judgment and takes into account the patient’s appearance and behavior. Although clinical judgment is a skill doctors use every day, she said it is hard to measure. The Phase II study now underway includes an electronic form designed to standardize such assessments. The form, which is part of the electronic medical record, asks physicians to rate the appearance of each cancer patient being admitted for fever and neutropenia as well, sick or toxic. Toxic includes children who are non-responsive, not breathing or extremely ill.

    Gaur credited the Clinical Informatics department for creating the online form that provides clinicians an easy way to document the information and demonstrates the potential application of electronic medical records for clinical research.

    The current study also assesses the predictive value of two inflammatory markers. The markers are procalcitonin, which is an inactive peptide considered a possible sign of bacterial infection, and C - reactive protein, which is produced in the liver in response to infection or inflammation.

    Other authors of this study are Patricia Flynn, MD, and Katherine Knapp, MD, both of Infectious Diseases; Deo Kumar Srivastava, PhD; Chenghong Li, PhD; and James Okuma, all of Biostatistics.

    The research was supported in part by the National Institutes of Health and ALSAC.

    December 2009