Systematic review of risk prediction rules in febrile neutropenic episodes in children and young people undergoing treatment for malignant disease


This systematic review aimed to identify, critically appraise and synthesise evidence on the discriminatory ability and predictive accuracy of existing clinical decision models of risk stratification in febrile neutropenic episodes in children and young people undergoing treatment for malignant disease.


The systematic review of CDRs initially identified 2057 potential studies and finally included 24, of which 21 provided data in a usable format. It showed two groups of studies have been undertaken to risk-stratify children who present with febrile neutropenia. The first group examined the use of clinical examination to predict radiographic pneumonia (4 studies), the second group examined more general infectious complications (20 studies).

Quantitative pooling of the three higher quality ‘pneumonia’ studies using a standard random-effects model produced imprecise estimates of sensitivity 75% (95% CI 56.4% to 93.6%) and specificity (point estimate 67.9%; 95% CI 55.9% to 79.9%). The implications of these results are that for populations with a similar prevalence of pneumonia (~5%), the absence of signs or symptoms of infection of clinical examination produces a post-test probability of pneumonia of about 1.5%. This can justify the routine withholding of chest radiographs, but the clinician must remember that a number of children will have an occult pneumonia and chest X-rays in an unresolving fever may be fruitful despite an absence of signs. This arises not because of the accuracy of the test, but the low prevalence of the condition.

The studies examining general infectious complications produced 16 separate models, and contained nine datasets used to validate previously derived models. They studied a variety of outcomes, with individual differences in definitions, but covered five main categories: death, critical care requirement, serious medical complication, significant bacterial infection, and bacteraemia. The performance of only one rule could be reasonably assessed across multiple datasets, that of AMC/Temperature criteria proposed by Rackoff [45] to exclude bacteraemia. This model, being tested by different groups across time and in different centres, has the greatest strength of evidence. The most appropriate pooled estimate of the rule’s effectiveness comes from a random effects model assuming no threshold variability, and excluding both the derivation sample and an outlying study. This led to estimates of moderate discriminatory ability (LR [low] = 0.22 (95% CrI 0.03 to 1.85), LR [medium] = 0.79 (95% CrI 0.12 to 2.06), LR [high] = 3.41 (95% CrI 0.24 to 18.7)).

Of the other rules the Santolaya model shows a good ability to differentiate between low- and high- risks groups when considering a wider definition of ‘serious infection’, (LR [low] = 0.17 (95% CI 0.12 to 0.23) LR [high] = 2.87 (95% CI 2.43 to 3.38) ). The rule has been developed and tested in Chile, which may limit its applicability in Western Europe and North America. The proportion of patients with bacteraemia (~25%) is similar to the other studies in this review, but their broad definition of adverse medical outcomes are found in ~60% of cases does not have a direct comparator among the Western European / North American studies reviewed and no accurate conclusion can be reached. Other rules show promise and have clinical/physiological similarities, but have not undergone extensive testing.

An exploratory analysis of the individual features common across predictive studies show, age, malignant disease state, clinical assessments of circulatory and respiratory compromise, higher temperatures and bone marrow suppression all have some explanatory power. Any adaptation or development of a new rule should primarily look to assess these variables over those found purely by ‘p-value’ sampling of bivariable testing. Differences between the individual predictive ability of variables and the specific outcomes chosen have some clinical plausibility; for example to failure of local infection to predict bacteraemia may be reasonable as infection that has been isolated by the individual’s immune system has a reduced opportunity to seed into the bloodstream, yet may still cause significant problems requiring intensive antibiotic therapy as so be recorded as a significant infectious complication.

Conducted by: Bob Phillips1, Ros Wade1, Lesley Stewart1, Alex J Sutton2

1. Centre for Reviews and Dissemination; 2. Department of Health Sciences, University of Leicester


Phillips B, Wade R , Stewart LA, Sutton AJ. Systematic review and meta-analysis of the discriminatory performance of risk prediction rules in febrile neutropaenic episodes in children and young people. Eur J Cancer 2010;46(16):2950-2964


Commissioned by the Medical Research Council