Pediatric cancer patients have an increased risk of potentially life-threatening infectious [1–4] complications due to their underlying illnesses and intensive anticancer treatment [5–7]. Most of these infections are bloodstream infections (BSIs), most often associated with a central venous access device (CVAD) [8]. Healthcare-associated infections (HAIs) in pediatric cancer patients have been investigated in only a few single-center studies [9–12]. One additional European study included two pediatric oncology units [13] in a 6-month prospective survey using periodic chart review during hospitalization, but did not use adjusted definitions for immunocompromised pediatric patients [14].

The exact proportion of HAIs in these high-risk patients is unknown; two single-center prospective studies in pediatric hematology-oncology patients, excluding allogenic bone marrow transplantations, found an incidence of 20% [9] (ID 10.8/1,000 inpatient days) and 24% [10] (ID 17.7/1,000 inpatient days), respectively. One recent single-center study found an ID of 38.9 HAIs per 1,000 inpatient days in a pediatric stem-cell and bone-marrow transplantation unit [12], but the corresponding protocol referred only to patients with neutropenia [15].

In 1998, our group developed a module for the prospective unit-based surveillance of HAI in pediatric cancer patients: Onkopaed NKI. The results of a prospective single-center 10-month pilot study confirmed the feasibility of this module as well as the high risk of HAI and CVAD-associated BSIs in pediatric cancer patients [9]. The Onkopaed NKI module has already been used for prospective intervention studies on CVAD-associated infections [16], the restricted use of glycopeptides as a component of empirical antimicrobial therapy [17], and antifungal prophylaxis with itraconazole [18]. We now present the results of the multicenter prospective surveillance study for HAI and nosocomial fever of unknown origin (nFUO) in 7 German pediatric oncology centers from 2001 through 2005.

The Onkopaed NKI module [9] uses standard methods of the Centers for Disease Control and Prevention (CDC; Atlanta, GA) [19] for the prospective unit-based surveillance of selected HAIs and nFUOs (occurring at least 24 hours after patients' admission) in pediatric cancer patients. The most important outcome parameter, the ID of events per 1,000 inpatient days, was calculated as the number of events divided by all inpatient days times 1,000. The ID of events per 1,000 inpatient device utilization days and the device utilization rate itself were calculated correspondingly. 'Multidrug-resistant isolates' referred to bacterial pathogens in vitro resistant to at least two of the corresponding first-line antimicrobials (ceftazidime, ceftriaxone, aminoglycosides, piperacillin-tazobactam). The surveillance module did not ask for admissions or numbers of individual patients in the participating institution during the surveillance period.

This study is the first prospective multicenter surveillance study of HAI in pediatric cancer patients using standardized CDC methods and adjusted consensus definitions for this special patient population. Our study confirms that pediatric cancer patients are at an increased risk for specific HAI. These adverse events can result not only in significant morbidity and mortality, but also in an increased expenditure of limited financial and personnel resources. In particular, CVAD-associated BSI, which represent the majority of all HAIs in this population, should be targeted to improve clinical practices and evaluate new preventive strategies [23].

Although associated with a severe clinical course, a significant proportion of all patients with HAI did not have neutropenia at the time of diagnosis. This is in agreement with results from other groups who found severe CVAD-related infections in pediatric cancer patients in the absence of neutropenia [24]. As a consequence, surveillance should not be limited to the period of neutropenia, and for the calculation of IDs days of neutropenia should not be used as the (sole) denominator.

Even in those patients who have received an allogenic BMT or SCT, the clinical relevance of certain late onset infections (BSIs or invasive fungal infections after the recovery of the leukocyte count have been observed) [7] argues against a surveillance that is restricted to early 'neutropenic' HAI.

The IDs described in our study are within the range of those from single-center studies. During 3 months of surveillance in 52 pediatric cancer patients with 135 admissions and 1,017 inpatient days, Urrea et al. [10] found a HAI rate of 23.5%, an incidence of 13.3/100 admissions and an ID of 17.7 per 1,000 patient days. In the pilot study of the Oncopaed NKI module [9], we investigated 40 HAIs in 28 out of 143 patients (HAI rate 19.6%) hospitalized for 3,701 days (776 admissions) during a 10-month prospective surveillance period. The HAI incidence was 5.2/100 admissions and the overall ID of HAI was 10.8 per 1,000 inpatient-days. The overall CVAD-related BSI rate was 7.4 per 1,000 inpatient utilization days.

In the transplantation unit (C3), patients after an autologous STC had the same ID for HAI as those after allogenic transplantation. This is in line with a prospective surveillance study in pediatric patients treated in an STC unit conducted by Laws et al. [12]. Using the surveillance protocol of the German National Reference Centre for Surveillance of HAI for adults [15], 34 pediatric and 39 adult patients were followed for 24 months. The incidence of HAI during neutropenia (1,156 neutropenic days) was 38.9 per 1,000 days, without significant difference between the groups. The Oncopaed NKI module may be used for the surveillance of late onset infections after SCT, in particular invasive aspergillosis [25].

Overall, the attributable mortality in our study was surprisingly low, mainly due to a low mortality in patients with BSIs. The outcome of BSIs in pediatric oncology patients has improved over the last decade, mainly due to early aggressive treatment in any suspected case of sepsis and due to advances in intensive care in this particular population [26].

The proportion of resistant isolates in the BSIs was also surprisingly low, considering the increasing proportion of resistant isolates in studies from other countries [13].

Wisplinghof et al. [27] determined antimicrobial susceptibilities of nosocomial BSI isolates in pediatric patients in the US SCOPE project. As in our study, Gram-positive pathogens accounted for most (65%) of the cases, Gram-negative pathogens for 24% and fungi for 11%. We identified only a single case of candidemia. In Wisplinghof's study, the proportion of MRSA increased from 10% in 1995 to 29% in 2001 [27]. Vancomycin-resistance was seen in 11% of E. faecium isolates (VRE). In our study, no BSI was caused by MRSA, but 2 of 5 E. faecium BSI were caused by VRE. In the study of Raymond et al. [13], the prevalence of antimicrobial resistance depended on the type of unit. Up to 26% of S. aureus and 89% of CoNS were methicillin-resistant (MRSA, MRSE), and 38% of K. pneumoniae expressed an extended-spectrum beta-lactamase (ESBL) [28].

The high mortality in patients with invasive aspergillosis, in particular in those patients with disseminated disease, is still a matter of great concern. The recognition of the role of external reservoirs and acute exposure to fungal spores as well as the expanding scope of the agents available for antifungal prophylaxis may contribute to a higher success rate in the prevention of these often fatal infections in high-risk pediatric cancer patients [18].

Based on our experience with a norovirus outbreak in pediatric oncology patients [29] and our recent investigation of rotavirus infections in a matched pairs analysis [30], the surveillance of gastrointestinal HAI in pediatric cancer patients should include at least rotavirus, norovirus, and CDAD, which is, however, difficult to define in this population [31].

Our results confirm the observation of Dettenkofer et al. [15] that a remarkable proportion of all symptomatic patients with CDAD do not suffer from neutropenia at the time of diagnosis. The empirical antimicrobial regimen of fever and neutropenia does not cover C. difficile and fosters the production of C. difficile toxins [32]. Discontinuation of the offending antibiotic as a general treatment principle in patients with CDAD [33] is not feasible in patients with neutropenia. Recently, hypervirulent nosocomially transmitted C. difficile isolates have been described in several countries, causing increased morbidity, length of hospital stay, and mortality in adult patients [34].

Considering the significant morbidity of affected patients, the possibility of outbreaks, and the emerging threat of hypervirulent isolates, it seems mandatory to implement continuous prospective surveillance of CDAD in pediatric cancer patients [35].

Prospective surveillance modules, such as that presented here, may be utilized to detect early (< 30 days) or late (up to 2 years in presence of prosthetic devices) SSI. Gaur et al. [36] retrospectively investigated infections complicating limb-sparing surgery (104 procedures) in 103 children and young adults with bone malignancies. They found a high incidence of infections (median 4%; range 0–13%), including local SSI in 67% and secondary bacteremia in 21% of patients. Patients who developed deep infections were more likely to undergo amputation (odds ratio [OR] 24.0; 95% CI 5.1–114.0; P < 0.001) and were less likely to have good functional outcomes (OR 0.02; 95%CI 0.002–0.15; P < 0.001). Different strategies of antimicrobial prophylaxis and antibacterial management in tumor-related orthopedic surgery may – as well as in other surgical interventions [37] – be evaluated prospectively in future studies [38].

The fact that we defined as nosocomially acquired infection only those RSV infections in which the first symptoms occurred after at least 5 days of inpatient care did not allow the inclusion of some cases of nosocomial RSV infection, since the exact onset is sometimes difficult to determine. Although RSV is easily communicable and has been the cause of fatal outbreaks even in the setting of allogenic BMT [39], there is still no evidence-based antiviral treatment option [40]. Therefore, the early detection of this pathogen and immediate contact/droplet isolation of infected patients and their relatives is of utmost importance.

Our results demonstrate that microbiologic testing in pediatric cancer patients with pneumonia, with the exception of mechanically ventilated patients with acute respiratory failure, often fails to identify the causative agent [6]. In the light of the profoundly restricted time budget of infection control personnel, the inclusion of pneumonias without a confirmed pathogen, blood-culture negative BSIs, urinary tract infections, and nFUOs should be reconsidered critically in surveillance efforts in this population.

While we did not control for confounding variables such as demographic characteristics of the patients, duration of neutropenia or illness severity, further studies are needed to confirm that participation in such a surveillance study results in a significant decrease in HAI rates. This was not an interventional study and each participating center decided on its own responsibility about any practical consequence related to the reported HAI rates.

Today, HAIs are considered to be important adverse outcomes in terms of increased morbidity and mortality, postponed chemotherapy treatment cycles, as well as prolonged duration of hospital stay and additional costs from the perspective of the caregiver. All efforts to prevent HAI and to contain the nosocomial spread of pathogens are critical aspects of patient's safety in the hospital environment [41]. Thus, the prospective surveillance of HAI and comparison with standardized rates should be recognized as a fundamental component of quality assurance and control in the treatment of immunocompromized pediatric cancer patients.

Our multicenter study confirmed that pediatric cancer patients are at an increased risk for specific HAIs, in particular vascular catheter-associated bloodstream infections. It seems reasonable to implement a specific surveillance protocol for HAI in this high risk population. Surveillance efforts should not only focus on periods of profound neutropenia. The prospective surveillance of HAI and comparison with cumulative multicenter results are indispensable for targeted prevention of these adverse events of anticancer treatment.