paediatrics Brussels 17

294

Kennedy et al.

International Journal of Radiation Oncology Biology Physics

Results

age-appropriate booklets of questionnaires to eligible participants. The study was approved by ethics committees in all participating countries.

Baseline characteristics

Outcome measures

Outcomes were ascertained in 151 of 244 eligible survivors (61.9%) (74 of 117 [63.2%] and 77 of 127 [60.6%] that had received HFRT and STRT, respectively) at a median interval from diagnosis of 5.8 (range, 4.2-9.9) years. Participants and non- participants in the QoS study receiving HFRT were similar with respect to sex, age at diagnosis, pre- and postoperative charac- teristics, chemotherapy received, and interval between diagnosis and assessment of QoS to those receiving STRT; there was a small excess of neurologic deficits of extraocular movement in those allocated to HFRT and of premorbid developmental impairment in those allocated to STRT ( Table 1 ). The radiation therapy actually delivered to PNET4 QoS study participants corresponded well to that prescribed in both treatment arms: quality assurance included both fields (checked in at least 68 of 151 [45%]) and dose delivered (checked in at least 51 of 151 [34%]). There were significant ( P < .01) correlations between all of the outcome measures, including strong positive correlations (0.58- 0.80) between proxy- and self-report scores, between health status and HRQoL scores, and between executive function and behav- ioral difficulties scores ( Supplementary Table e1 , available on- line). Female participants had poorer HRQoL z -scores than males (group mean difference 0.48, 95% confidence interval [CI] 0.13- 0.84, P Z .008) but were similar to males with respect to executive function, health status, and behavioral difficulties. Scores in subgroups governed by questionnaire age ranges showed no statistically significant differences between treatment arms other than lower (better) adult self-report BRIEF executive func- tion scores in those aged 18 years at assessment ( Fig. 1 ). In the merged single dataset of z -scores for all participants for the 4 principal outcome measures, executive function (BRIEF) z -scores were significantly lower (better) after HFRT than after STRT (group mean difference 0.48, 95% CI 0.16-0.81, P Z .004) ( Table 2 ), but health status, behavioral difficulties, and HRQoL of all participants were similar in the 2 treatment arms (group mean z -score differences 0.26, 0.23, and 0.14, respectively, with 95% CIs including the null point, P .25) ( Table 2 ). The intergroup differences in executive function (BRIEF) Global Executive Composite z -scores seemed to be general effects reflected in sta- tistically significant intergroup differences for both the Behavioral Regulation Index, carried by inhibition, shift, and emotional control subscales, and also for the Meta-cognitive Index, carried by monitoring, working memory, and planning/organizing subscores ( Supplementary Table e2 ). Outcomes at posttreatment evaluation Effect of HFRT on executive function, behavior, health status, and HRQoL

The 4 principal QoS outcome measures were z -scores on stan- dardized age-appropriate questionnaires. These assessed executive function in everyday life, health status, behavioral difficulties, and HRQoL. In participants aged < 18 years at assessment, the ques- tionnaires were parent-report versions of the Behavior Rating Inventory of Executive Function (BRIEF) (33) ; the Health Utili- ties Index (HUI3) (34) ; the Strengths and Difficulties Question- naire (SDQ) (35) ; the Pediatric Quality of Life Inventory (PedsQL) (36) ; and, if aged 11-17 years, self-report versions of the HUI3, SDQ, and PedsQL. Survivors aged 18 years provided self-report versions of the BRIEF, the HUI3, and the 30 core question version of the European Organization for Research and Treatment of Cancer Quality of Life measure (QLQ-C30) (37) . Reduction in height z -score compared with that at diagnosis, known to vary with radiation treatment dose (38) , was the pre- specified principal endocrine outcome. Height, weight, and mid-parental height were expressed as z -scores, age- and sex- standardized against United Kingdom British 1990 growth refer- ence values (39) to allow inter- and intragroup comparisons. Baseline demographic and secondary outcome information was provided by clinicians using adapted versions of the Medical Examination Form (40) and by adult participants and parents of child participants using the Medical, Educational, Employment and Social questionnaire (40) . Univariate analyses of HFRT and STRT group scores were con- ducted using Mann-Whitney U tests or t tests. Quality of survival questionnaire scores in age-specific subgroups, governed by the age ranges of questionnaires, were converted to a single dataset of z -scores in all participants where mean Z 0, SD Z 1 for scores of all study participants on that measure. Proxy-report z -scores for children and self-report z -scores for adults for BRIEF and HUI3 were thus used to create a single “executive function z -score” variable and a single “health status z -score” variable, respectively. Proxy-report SDQ z -scores, available only in participants aged < 18 years, provided behavioral difficulties z -scores. Finally, child self-report z -scores for PedsQL and adult self-report z -scores for the QLQ-C30 were used to create a single “HRQoL z -score” variable. Statistical analysis of QoS was thus simplified into 4 analyses relating to these 4 prespecified principal QoS outcomes. Analyses of questionnaire subscales were conducted only when total scores differed. A regression model, including sex, younger age (3-7.9 years) or older age (8.0-20.8 years) at diagnosis, and cerebellar mutism (or, in an alternative model, perioperative complications including cerebellar mutism), was used to increase the precision of the estimate of the effect of treatment on principal QoS outcomes and on decrement since diagnosis in height z -score. Sensitivity ana- lyses were used to examine possible confounding by baseline characteristics. Statistical significance testing (SPSS version 19.0; SPSS, Chicago, IL) was 2-tailed with a 1% significance level to adjust for multiple testing (41) . Statistical analysis

Effect of HFRT on growth

Compared with STRT, the mean group decrement since diagnosis in height z -score after HFRT was greater by 0.43 (95% CI