Examining health-related quality of life in pediatric cancer patients with febrile neutropenia: Factors predicting poor recovery in children and their parents
The impact febrile neutropenia (FN) has on the health-related quality of life (HRQoL) of children with cancer and their families is poorly understood. We sought to characterize the course of child and parent HRQoL during and following FN episodes.
Data on HRQoL were collected in the multisite Australian Predicting Infectious ComplicatioNs in Children with Cancer (PICNICC) study. Participants were enrolled between November 2016 to January 2018. The Child Health Utility (CHU9D) was used to assess HRQoL in children (N = 167 FN events) and the Assessment of Quality of Life (AQoL-8D) was used to assess HRQoL parents (N = 218 FN events) at three time points: 0–3 days, 7-days and 30-days following the onset of FN. Group-based trajectory modeling (GBTM) was used to characterize the course of HRQoL.
For children, three distinct groups were identified: persistently low HRQoL over the 30-day course of follow-up (chronic: N = 78/167; 47%), increasing HRQoL after the onset of FN to 30 days follow-up (recovering: N = 36/167; 22%), and persistently high HRQoL at all three timepoints (resilient: N = 53/167; 32%). Applying these definitions, parents were classified into two distinct groups: chronic (N = 107/218, 49%) and resilient (N = 111/218, 51%). The child being male, having solid cancer, the presence of financial stress, and relationship difficulties between the parent and child were significant predictors of chronic group membership for both parents and children. Children classified as high-risk FN were significantly more likely to belong to the recovery group. Being female, having blood cancers and the absence of financial or relationship difficulties were predictive of both parents and children being in the resilient group.
Approximately half the children and parents had chronically low HRQoL scores, which did not improve following resolution of the FN episode. The child’s sex, cancer type, and presence of financial and relationship stress were predictive of chronic group membership for both parents and children. These families may benefit from increased financial and psychosocial support during anti-cancer treatment.
National Health and Medical Research Council Grant (APP1104527).
]. While a bacterial infection is documented in around 20% of children with FN and 3% require intensive-care-unit (ICU) level care, contemporary studies still report that over 50% of children recover quickly and do not have a clinical infection [
]. A risk-stratified approach to care should be considered in FN, enabling low-risk patients to be managed with reduced-intensity treatment, including oral antibiotics, with the option for administration in the comfort of their own home [
]. To our knowledge, the only study to assess HRQoL in children and their parents during an FN episode was Orme et al. 2014 [
]. However, this study assessed HRQoL using the visual analogue scale which has limited external comparability. Further, this study only assessed HRQoL during the FN episode but did not examine HRQoL after the FN episode resolved.
]. Moreover, characterizing and quantifying the HRQoL effects associated with FN is necessary to assess the cost effectiveness of FN care. This facilitates the assessment of the cost-effectiveness of existing or emerging approaches to the management of FN (in particular for low-risk FN) [
The Australian Predicting Infectious ComplicatioNs in Children with Cancer (PICNICC) study was a multi-site, prospective validation study that also captured comprehensive HRQoL and economic data on children with cancer and FN. Our study used HRQoL data from PICNICC to describe the course of HRQoL of children and their parents during and following an FN episode. We hypothesised that FN would have a temporary deleterious and heterogenous impact on HRQoL, which would dissipate following FN resolution. Therefore, this study sought to describe individual variation between children and their parents and to identify predictors of poor HRQoL.
2.1 Participants and procedures
]. Briefly, children with cancer on active treatment who were admitted to hospital or presented to emergency, outpatient or day-chemotherapy departments with fever or clinical instability were eligible for inclusion. Fever was defined as a temperature ≥38 °C and neutropenia was defined as an absolute neutrophil count (ANC) 3. Children with a hematopoietic stem cell transplant (HSCT) within three months were excluded. Multiple, discrete FN episodes per patient were allowed [
During the study period, children were managed according to hospital FN guidelines. Risk stratification or home-based treatment of low risk FN was not routinely used.
This study had primary approval from the Royal Children’s Hospital Melbourne Human Research Ethics Committee (RCH HREC 36040A) and was ratified by the University of Technology Sydney HREC ETH17–1128. Informed parent consent was obtained prior to the child’s enrolment in the study.
2.3 Main outcome measure and HRQoL instruments
]. Parent’s HRQoL was assessed using the Assessment of Quality of Life 8 Dimensions (AQoL-8D), a generic MAUI, which assesses HRQoL over the two super dimensions ‘physical health’ (independent living, senses, pain) and ‘mental health’ (self-worth, coping, relationships, happiness) [
HRQoL data were collected at three time points following FN onset: within 0–3 days (in hospital) and at 7-days (at discharge) and at 30-days. Children were included if they had completed the CHU9D questionnaire for at least one FN episode. Parents were included if they had completed the AQoL-8D questionnaire for at least of one of their child’s FN episodes. A complete case analysis was used.
]. Australian-specific utility values were applied to the completed questionnaires to measure HRQoL according to the protocols specified by the developers [
2.4 Statistical analysis
]. The GBTM is robust to serial correlation when applied to patients with multiple FN episodes and multiple courses of follow-up as it assumes conditional independence. That is for each individual within a given trajectory group, the distribution of the current observations of the outcome variable (yt) is independent of prior values (yt-1) [
The GBTM uses maximum likelihood to model the probability of a specified number of underlying latent trajectories and their shape. Statistical criteria for ascertaining the best fitting model included Bayesian Information Criterion (BIC). Group-specific average posterior probabilities and estimated group sizes were also used to calculate group-specific odds of correct classification to assess entropy. Other criteria included nonoverlapping confidence intervals (CIs), distinct average posterior probabilities across groups, and whether the additional trajectory added clinically relevant information (see Table S8 and Table S9).
Statistical analyses were performed using STATA/SE 16 and the traj Plugin for STATA (StataCorp, College Station, TX). A censored normal distribution (minimum = −0.10, maximum =1) was used in the model and applied to HRQoL values for both parent and child. Univariate associations between predictors and HRQoL trajectories were examined using trajectory risk factor analyses. In addition, a multivariate model including all predictive factors was investigated. All variables found to be significant in univariate analyses at p-value ≤0.05 were included in subsequent multivariate analyses. Age, sex, and time with cancer were included in multivariate analyses regardless of significance as they were considered important control variables. To account for multiple FN episodes per patient, multinomial and logistic regression with cluster robust standard error were used (see Table S7). A subsequent exploratory analysis examined the linear association between these risk factors and parent and child HRQoL using mixed-effects linear regression (see Table S6).
2.5 Risk factors
], time to first antibiotic dose after hospital presentation, chemotherapy intensity (more or less intensive than acute lymphoblastic leukemia (ALL) maintenance), and cancer diagnosis were reported as part of the PICNICC variables. Cancer diagnosis was dichotomously characterised as “blood cancer” (i.e. lymphoma/leukemia) or “solid cancer” (i.e. sarcoma, brain tumours or other solid tumours). The presence of financial stress and relationship distress between the parent and child was assessed using two items from the Care-related Quality of Life instrument (CarerQol) questionnaire [
]: (1) I have financial problems due to my child’s cancer and (2) I have relationship problems with my child. Responses were dichotomised as “no” and “some/a lot” (herein referred to as relationship strain).
2.6 Role of funding source
This study was funded by National Health and Medical Research Council (NHMRC) Project Grant (APP1104527). The NHMRC was not involved in study design, data collection, analysis or manuscript preparation or approval. The corresponding author (Anna Crothers), and her supervisor (Richard De Abreu Lourenco) had full access to all the data in the study and had final responsibility for the decision to submit for publication.
]. Further, this is the first time the CHU9D has been used to assess HRQoL in a pediatric cancer population. Our results show that the onset of an FN episode had a significant and heterogeneous impact on the HRQoL of children with cancer and their parents. For children and parents in the resilient and recovery trajectory groups, the onset of FN had a temporary deleterious impact on HRQoL, which dissipated following FN resolution. However, for children and parents in the chronic groups, there was no improvement following FN resolution suggesting the low HRQoL may be unrelated to the FN episode. In support of this hypothesis, and consistent across parent and child models, time-invariant factors predictive of chronic group membership were the child being male, the child having solid cancer or undergoing intensive chemotherapy, experiencing financial stress and the presence of relationship difficulties between parent and child.
]. This allowed us to identify subpopulations of children and parents with chronically low HRQoL scores, which were significantly below population norms [
], at both the onset of an FN episode and up to 30-days after. We were also able to identify children and parents with HRQoL scores that quickly recovered following the resolution of the FN episode (i.e. ‘resilient’ and ‘recovery’ groups). In understanding the existence of such trajectory groups and the characteristics which differentiated them, it may be possible to identify ‘at risk’ subpopulations with early divergent trajectories as potential intervention targets [
] and up to five years post-treatment [
]. Intensive chemotherapy regimens require more frequent hospital visits and hospitalization and are associated with more side effects such as mucositis, pain or infective episodes. Furthermore, the prognosis of many solid tumours is worse than for ALL . A strong association between financial distress and reduced HRQoL in both children and their parents was also evident. Moreover, 80% of parents reported financial stress due to their child’s cancer. This was despite Australia having a dedicated income support payment for carers of children with chronic illness as well as additional support available from some cancer charities . Previous research has found a strong correlation between financial stress and reduced HRQoL in adult cancer patients [
]. Our results suggest that this impact extends to pediatric cancer patients. Families reporting financial distress have higher odds of experiencing worse HRQoL that persists despite the resolution of the FN episode. These are important finding given that a validated tool exists to identify those experiencing financial distress [
]. Not surprisingly, we also found that families in which there is some degree of relationship distress between parents and children were significantly more likely to be assigned to the chronic group. Identifying families experiencing financial or relationship stress at the commencement of cancer treatment may enable targeted interventions to support both the child and parents throughout the treatment course.
]. This has led to the development of important treatment pathways, focusing on the early discharge to home-based care of patients who are at low risk for serious infection [
]. In keeping with this finding, our results did find a significant association between increased LOS and antibiotic duration, and increased odds of chronic group membership. In contrast, results of our trajectory modeling did not find an association between receipt of home-based care via Hospital-in-the-home programs and HRQoL, for either parents or children (Table 2 and Table S7). Of note, formal low-risk FN programs were not implemented during this study, and the home-based care was likely due to other reasons such as treatment of established infections or other cancer-related complications. Discrete choice experiments that have examined treatment preferences of parents with children undergoing active cancer treatment have found that many parents still prefer inpatient management over outpatient management for low-risk FN episodes [
], whilst parents who valued HRQoL were more likely to opt for outpatient care [
]. More recently, a clear preference for home-based treatment of low-risk FN was identified by clinicians and parents due to perceived improvements in HRQoL [
]. This highlights the importance of consumer engagement when home-based FN care pathways are developed and implemented to ensure that parent and patient concerns and reservations are appropriately addressed.
]. A randomised controlled trial, comparing inpatient and outpatient treatment of pediatric FN that was not included in this review, did show that outpatient management of low-risk FN provided significant benefit to parents and patients across several HRQoL domains [
]. Our study is, however, the largest study of its kind and provides a comprehensive baseline understanding of the HRQoL impact of standard inpatient management of FN. It is also the first to quantify pediatric and parent HRQoL associated with FN using the validated CHU9D and AQoL-8D tools, and to do so in a manner that can be directly incorporated into economic evaluations. We present results for HRQoL over an acute period (0–7 days after FN onset) and after the resolution of the FN episode (7–30 days after FN onset). This 30-day follow-up period reflects the expected period for resolution and recovery from an FN event [
]. However, as we are the first study to examine the HRQoL of children with cancer using the CHU9D, it was not possible to determine whether children in our study fully recovered from their FN episode at 30-days follow-up by benchmarking against other CHU9D results from other studies in which children were not experiencing FN. Further, as our study only captured information on cancer treatment prior to the FN event but not during 30-day follow-up period, it is unknown if children had recommenced chemotherapy. This is particularly important for children with solid cancers who typically receive cancer treatments in 21-day cycles.
Febrile neutropenia has a significant and heterogeneous impact on the HRQoL of children with cancer and their parents in resilient and recovery trajectory groups. However, for children and parents in the chronic group, their persistently low HRQoL remained static throughout. Risk factors for chronic group members include the child being male, having solid cancer, undergoing intensive chemotherapy, financial stress and the presence of relationship difficulties between parent and child. These factors could be used to identify children and families who would benefit from additional financial and psychological support during their cancer care. Combined with our improved understanding of pediatric FN, our results provide critical insights for the current and future management of pediatric cancer patients and their families.
Declaration of Competing Interest
DZ reports personal fees from Bayer, personal fees from Amgen and personal fees from Day One, outside the submitted work. AC reports grants from NHMRC APP1104527, during the conduct of the study. GMH reports grants from NHMRC APP1104527 and the Victorian Cancer Agency early career fellowship, during the conduct of the study. MAS reports grants from Merck, from Gilead Sciences, and from F2G, personal fees from Pfizer and personal fees from Gilead Sciences, outside the submitted work. RDAL reports grants from NHMRC APP1104527, during the conduct of the study. RP reports grants from NIHR, during the conduct of the study. All the other authors report no conflicts.