Prospective evaluation of quality of life in children treated in UKALL 2003 for acute lymphoblastic leukaemia: A cohort study

Health‐related quality of life (HRQoL) from diagnosis until end of treatment for children with acute lymphoblastic leukaemia was investigated, examining effects of age, gender, risk‐stratified treatment regimen, and therapy intensity (one vs. two ‘delayed intensifications’ [DIs]).

classified as 'MRD low-risk' (undetectable MRD at induction day 29 or detectable <0.001% leukaemic cells at day 29 becoming undetectable by week 11) were randomly assigned to one (experimental arm) or two (standard therapy) blocks of 'delayed intensification' (DI) chemotherapy prior to maintenance. This additional chemotherapy comprised doxorubicin, etoposide, cyclophosphamide over a 5-day period followed by a period of neutropenia, with blood count recovery by day 21. Patients with MRD ≥ 0.001% at the end of induction were classified as 'MRD high-risk' and were randomised to continue standard therapy (Regimen A or B) or to intensify treatment further (Regimen C). There was an improvement in 5-year event-free survival to 87% in the trial overall, with no increase in relapse risk associated with de-escalation of treatment in the low-risk group 2 and a reduction in relapse risk in high-risk patients who received more intensive treatment. 3 These excellent survival outcomes raise questions about how to balance treatment-related morbidity and health-related quality of life (HRQoL). Physical side effects of chemotherapy, 4 repeated hospitalisations and associated limitations for social and physical opportunities 5 compromise a child's HRQoL.
When assessing HRQoL, a distinction is made between generic and disease-specific measures. 6 Generic measures enable comparison with the general population, while disease-specific measures focus on disease symptoms, and are considered more sensitive to evaluate different treatments. Although ratings of child HRQoL should be made by both child and parent, 7 it is often necessary to rely on parents' proxy ratings, especially where children are too young or ill to respond themselves. Parents are under intense stress given the emotional and financial costs of caring for a sick child. 8 Family care-giving burden therefore needs to be considered as an integral part of any comprehensive evaluation of HRQoL.
The aim of this study was to assess HRQoL of children treated in the UKALL 2003 trial from diagnosis until end of treatment. Parent proxy and patient reports (from children aged >8) were collected. Specifically, the following were determined: (i) generic HRQoL compared with population norms; (ii) changes in generic and disease-specific HRQoL over time, depending on child age, gender and treatment regimen; (iii) differences in HRQoL between low-risk patients randomised to treatment reduction and those receiving standard care; (iv) differences in HRQoL between high-risk patients randomised to treatment intensification and those receiving standard care.

HRQoL study
Questionnaires were completed at five time points (time points 1-5, T1-T5) during scheduled clinic appointments. T1: As soon as possible after diagnosis, with parents asked to provide a baseline assessment of their child's HRQoL before diagnosis (and care-giving burden immediately after diagnosis); T2: (week 4) end of induction chemotherapy; T3: immediately prior to maintenance therapy (week 23 for patients who received Regimen A and one DI to week 47 for those who received Regimen C); T4: completed at 18 months (during maintenance chemotherapy); T5: end of therapy.

Child's HRQoL
The PedsQL4.0 generic core 9 is a 23-item scale that yields three scores about side effects, cognitive problems, concern for appearance and communication. In order to ensure relevance to a U.K. sample, the Anglicised version of PedsQL which has been confirmed to be both valid and reliable in the U.K. population 12 was used. Questions which were also present in the PedsQL generic version were removed from the cancer module in order to reduce the overall length of the questionnaire. This resulted in a 19-item scale assessing pain and hurt, nausea, procedural anxiety, worry about side effects, concern for appearance and communication. Five-point response scales (0 = 'never a problem' to 4 = 'almost always a problem') were used, with higher scores representing worse outcomes.

Parental care-giving burden
A modified measure of parents' perceived care-giving burden in families with a child with asthma was used. 13

Statistical analysis
Responses were checked for temporal consistency (that response date fell close to scheduled time point (allowing for treatment schedules at T3, and gender at T5). Responses between 2 weeks before and 6 weeks after scheduled T1 or T2, and between 3 months before and 20 weeks after scheduled T3, T4 or T5 were considered acceptable. Responses outside these ranges were excluded as were those completed after relapse or stem cell transplant. These relatively large ranges were necessary given travel times to hospitals and differences in duration between treatment regimens.
HRQoL, demographics and treatment regimens grouped by response pattern over time were measured in order to compare nonresponders with responders and detect bias related to non-response.
Phi correlations and chi-square tests were used to determine relationships between non-response at one or more time points and gender, age, initial white blood count or treatment regimen. One-sample  Responses were compared for each of the 10 outcome variables between those who responded close to the due date (within 2 weeks of T1 and T2, and 6 weeks of T3, T4 and T5) with those who responded later, but within the required time frame (numbers were too small to compare those that responded earlier than expected). No differences were found between these groups for any generic or cancer-specific HRQoL subscales except that at T1, later respondents reported higher care-giving burden (mean = 3.23) than early respondents (mean = 2.64, P < 0.005) and at T2, later respondents reported less problem with nausea (mean = 0.97) than earlier respondents (mean = 1.41, Mean scores for HRQoL subscales across time, and comparisons with norms for healthy children are shown in Table 2. At each time point, Total, Physical and Psychosocial HRQoL scores were significantly lower than norms for healthy children.  Fig. 2A).

HRQoL change over time
Four cancer-specific HRQoL subscales (pain and hurt, procedural anxiety, communicating about illness and worries about side effects) showed a significant reduction between T2 and T3 before levelling between T3 and T5. Nausea increased significantly between T2 and T3 and then declined to T5. Concern for appearance showed no significant change over time (Fig. 2B). Care-giving burden was highest at T1 and T2, followed by a rapid then gradual decrease in scores reflecting a reduction in parental burden of care over time (Fig. 2C). There were few gender effects, except that parents of girls consistently reported worse problems with 'concern for appearance' than parents of boys (F = 29.08, P < 0.005). Effects of gender on change over time (i.e. gender by time-point interaction) were only significant for nausea (F = 7.33, P < 0.005). There was a significant increase in nausea between T2 and T3 in boys (P < 0.005) but the increase was smaller and non-significant in girls. Conversely, nausea decreased significantly between T4 and T5 in girls (P < 0.005) but less in boys (P = NS).

Effects of age, gender and treatment on HRQoL and change over time
Parents of older children were more likely to report that their child worried about side effects (F = 113.14, P < 0.005), and had concerns about appearance (F = 26.12, P < 0.005), but parents of younger children reported greater child procedural anxiety (F = 33.29, P < 0.005) especially at T2 (age by time-point interaction, F = 8.41, P < 0.005; age effect at T2, P < 0.005) and lessened with time.

DISCUSSION
This is a large prospective study of HRQoL in children and adolescents undergoing treatment for ALL, and it provides important outcome data charting the impact on child HRQoL. Children experience highly compromised HRQoL from diagnosis and up to 2 years later confirming previous cross-sectional findings. 15 This study supports the previous study findings that children have a significant reduction in HRQoL scores across all domains. 16 One advantage of this study is the prospective design, which allows measurement over the whole treatment course demonstrating changes in HRQoL scores over time. The study also identifies important differences between younger and older children in their reactions.
The study extends previous findings that show that children treated for ALL experience very compromised HRQoL immediately after diagnosis, 15 but HRQoL improves from 3 to 6 months and 1 year after diagnosis. 17 Parent reports of child HRQoL were much lower at T2 (4 weeks after diagnosis) than pre-treatment, and parental care-giving burden levels were at their highest at T1 and T2, confirming this initial period as one of great stress for families.
Few gender differences were identified, except that parents of girls consistently reported more problems with 'concern for appearance' . These concerns among girls may be amenable to direct support and intervention throughout the treatment period. There was greater increase in nausea between T2 and T3 in boys, although there is no clear reason for this.
Parents reported that symptoms such as nausea affected all children regardless of age, but older children were more concerned about side effects and appearance than younger children. Parents reported more procedural anxiety among younger children.  19 and these may overwhelm more subtle differences related to treatment.
There were a number of limitations to this study. Firstly, recruitment rates differed between centres, from 94% of eligible patients responding at least once to 18%. In addition, particularly at the later time points, there were a significant number of questionnaires that were not completed and this could adversely affect the validity of the results obtained. These differences may partly reflect availability of research staff between centres. No funding was available to support recruitment. Despite this, no differences in demographic or treatment were found between patients in the study and those not, suggesting that differential recruitment rates do not challenge the integrity of the findings.
One limitation of this study is that the assessment of HRQoL at diagnosis is a retrospective judgement and may therefore be unreliable.
However, the aim was to assess the impact of therapy on HRQoL and without this measure the baseline is entirely unknown. Many patients are unwell prior to diagnosis, and this may result in parents underestimating the child's pre-illness HRQoL.
The timing of recruitment around the five data collection time points was variable, especially T3 which was affected by allocated regiment (A, B or C), treatment delays and the number of DIs. T3 and T5 were chosen to represent particular points in the treatment protocol (pre-maintenance and end of treatment), rather than an absolute time point. Variation in response time was greater than anticipated and needs to be addressed in design of future similar trials. However, these data do not suggest that longer treatment for boys compromised their HRQoL more than girls in any measurable way, at least over the period of this study.
The age range of children recruited to the HRQoL study was 4-18 years, although the main trial included younger children and those up to 25 years of age. Difficulties were experienced obtaining responses from children, especially those in the younger age range, and no single HRQoL measure is sensitive across such a broad age range. As a result, younger children were not represented, even though they represent a significant proportion of the ALL population. This is an important F I G U R E 2 (A) Mean child generic (Total, Physical and Psychosocial) HRQoL by time (higher scores = higher functioning/QoL; (B) mean cancerspecific HRQoL subscale scores by time (higher scores = more problems); (C) mean parent's care-giving burden by time (higher score = higher burden) limitation, as studies have shown that there are clear differences between child self-reports and parent proxy report, particularly in social and emotional domains. There is generally greater concordance between responses to questions regarding physical functioning. 20,21 In addition, generic HRQoL was compared to U.S. population norms.
These norms were chosen given the extensive validation work that has been reported, 22 but may be sub-optimal for U.K. populations. The U.S. sample differs in age, sex and ethnic distribution compared to the UKALL 2003 sample, but given the way data were reported, it was impossible to allow for these differences in analysis. However, mean scores in the study population are so much lower than norms that these differences are unlikely to change the conclusions. The measure of care-giving burden was initially developed for work involving families of children with asthma. Although most items were relevant and the scale was acceptable to parents, a more specific care-giving measure might be more sensitive such as that developed by Wells et al. 23

CONCLUSION
This prospective study demonstrates a significant impact of therapy on HRQoL for children receiving treatment for ALL. Excellent survival rates mean that it is possible to reduce treatment intensity for some patients in an attempt to improve quality of life (QoL). It is therefore vital that accurate and sensitive QoL measurement is undertaken.
Further trials are needed to confirm these findings from the perspective of the patient and to determine whether HRQoL can be enhanced by clinical or supportive interventions for patients and their families A fuller understanding of the impact of therapy on HRQoL in patients with ALL will make an important contribution to the development of patient-reported outcomes among young people.