From January 2007 to May 2009, 73 children and adolescents aged from 2 years to 18 years with ITP were enrolled in this prospective study. They were recruited in seven pediatric centers of the region Pays de Loire, France. Patients were eligible if parental consent was obtained and if they did not have any concomitant disease that could impact their quality of life. If they were old enough, the child’s assent was also obtained. The study was approved by the comité de protection des personnes.

A single reference protocol was set up within each hospital, using a short-term corticosteroid therapy (4 mg/kg per day during 4 days, with a maximum dose of 180 mg per day) as the standard first-line treatment. The severity of ITP was stratified into four distinct groups, according to the platelet count and the existence of hemorrhagic signs, based on the consensus developed by the Intercontinental Childhood ITP Study Group (ICIS) (Table 1) 18 which we chose over the Buchanan score although the latter remains widely used 19 . Stages I and II do not require treatment and were subjected to clinical observation. Patients with stage III could either receive a 4-day course of corticosteroid or be subjected to simple surveillance, as judged appropriate by the treating physician. For stage IV, the first-line treatment was methylprednisolone or prednisone. Bone marrow aspiration was not required before initiation of corticosteroid therapy if an experienced practitioner could attest to the absence of signs of malignancy with normal blood cell smear. In cases of failure to respond to the initial corticosteroid treatment, or secondary recurrence, an infusion of intravenous immunoglobulin (IVIG) at 0.8 to 1g/kg was administered at 48-hour intervals if necessary. After diagnosis (D1), patients had outpatient follow-up visits at 1 month (M1), 3 months (M3) and 6 months (M6). Complete blood count, physical examination, adverse effects and ICIS bleeding scores were collected at each visit. Complete response was defined as a platelet count above 100.10⁹/L. If the patient relapsed, observation, corticosteroid therapy or IVIG could be considered according to the ICIS score, previous treatment, and the physician’s practice.

Quality of life was assessed by the PedsQL™ 4.0 Generic Core Scales. The PedsQL™ 4.0 Generic Core Scales is a generic questionnaire that measures HRQL by a self-report and/or a parent-proxy report for patients aged from 2 to 18 years: HRQL of patients under 5 years is exclusively measured by the parent proxy-report 20 . It was completed independently, at D1, M1, M3 and M6, by parents and children. D1 corresponds to the diagnosis day and also, generally, to the initiation of treatment. For younger children, an assistant was appointed to explain instructions and answer questions. Parent proxy-reports and age-specific child self-reports were similar versions, with formulations appropriate to the understanding of the responder. This questionnaire has been widely used in previous studies and its validity and reliability has been well demonstrated 20 21 22 .

We compared ITP children and parents mean PedsQL scores to previously published scores for the healthy reference population 20 21 23 . Independent sample t-tests were used to compare scores. Impaired HRQL was defined as a PedsQL score at least 1 standard deviation (SD) below the healthy sample mean score. Original norm data from a sample of healthy children was used as reference and scores from french healthy children were provided for information 21 22 23 24 .

The percentages of scores at the extremes of the score scale, the floor value (minimum possible score i.e. 0) and the ceiling value (maximum possible score i.e. 100), were described to explore the ability for ITP children to distinguish variations in quality of life. The ceiling effect is defined as an inability for children who had the maximum possible score to get a better score when a subsequent assessment is performed (the floor effect is the same phenomenon but with scores at the low end).

Intraclass Correlation coefficients (ICC) were used to describe agreement between child self-report and parent proxy-report across age subgroups. Intraclass Correlation Coefficients are designated as ≤ 0.40 for poor to fair agreement, 0.41–0.60 for moderate agreement, 0.61–0.80 for good agreement, and 0.81–1.00 for excellent agreement.

The minimal clinically important difference (MCID) concept was used to explore improvement in quality of life between diagnosis and M6. MCID is defined as the smallest difference in a PedsQL score of a domain of interest that each patient perceives to be beneficial and that would mandate a change in the patient’s management 21 . MCID scores were computed by using the Standard Error of Measurement (SEM) of scores at diagnosis. SEM was computed by multiplying the standard deviation by the square root of 1-alpha (Cronbach alpha coefficient). Excellent agreements between the SEM and MCID were shown in the literature 25 26 . Improvement in quality of life was defined as the score at M6 corresponding to the score at diagnosis with an improvement at least equal to the value of the MCID score.

Potential predicting factors that impact improvement in quality of life at M6 as previously defined were determined by univariate logistic regression. The associations are presented as beta coefficients with standard error (SE) derived from logistic regression. Given the exploratory nature of this analysis, we did not perform multivariate analyses. Statistical analyses were conducted using Stata software (version 12.1; StataCorp, Texas).

Between January 2007 and May 2009, 97 children were diagnosed with primary ITP in the region Pays de Loire. Among them, 16 children under 2 years of age were not eligible to be enrolled in the study because of age. Ultimately, 81 families were asked to complete the proxy and/or self-report version of the PedsQL™ 4.0 Generic Core Scales.

For the 34 children < 5 years, HRQL was exclusively measured by the parent proxy-report. Among the 47 self-reports of children > 5 years and the 81 parent-proxy reports, 8 reports in each group were partially uncompleted for more than 50% of the items. Therefore, according to the PedsQL™ 4.0 recommendations, these reports were considered as ineligible for the study. Finally, 73 children were included for analysis in the study: 34 aged 2 to 5 years, 39 aged 5 to 18 years, and 73 parents. Major characteristics of the 73 children at diagnosis are summarized in Table 2.

n: number of patients; IVIG: intravenous immunoglobulin;

*An observation unit is a unit of short hospital stay dedicated to diurnal monitoring.

**Although they were score I or II, some patients were not subjected to clinical observation and received treatment.

Initial treatments consisted in corticosteroids for 39 patients (53.5%) and IVIG for 9 patients (12.3%) while a wait-and-watch approach was used in 25 patients (34.2%). Between 3 and 6 months of follow up, 2 patients received IVIG only, 4 received corticosteroids only and 2 received IVIG and corticosteroids. After 6 months, platelet count was <100.10⁹/L in 26/69 patients (37%) including 11/42 patients under 10 years (26.1%) and 15/27 patients over 10 years (55.5%) (p = 0.1).

The main results concerning HRQL during the evolution of ITP are summarized in Table 3. The psycho-social summary score includes emotional, social and school functioning. Since we thought that change in scores between D1 and M6 were more suitable for assessing the evolution of HRQL, we mainly focused our analysis on these points.

% Floor/Ceiling = the proportion of scores at the extremes of the scaling range/SD = Standard deviation. Impaired health-related QOL was defined as a score that was more than 1 SD below the healthy sample mean/MCID = Minimal Clinically Important Difference. Standard Error of measurement computed by multiplying the standard deviation at diagnosis by the square root of 1 - alpha (Cronbach alpha coefficient)/Δ = the difference between mean scores at diagnosis and month 6. An improvement in quality of life between diagnosis and month 6 requires a Δ at least equal to the value of MCID score.

*p < 0.01.

¹Varni et al., 2007: correspond to the standards obtained from a population of 5480 healthy US children ().

²Tessier et al., 2009.

We compared self-reported and parent-proxy global, physical and psycho-social mean scores of our ITP population to the mean scores of the US population (original norm data) and the mean scores of the French population. At study onset, no difference was found regardless of the healthy population reference since scores observed in the population on study did not range below SD compared to the healthy population scores. However, when compared to the US population, we noted lower mean scores at diagnosis in patients with ITP for emotional functioning and school functioning (respectively 72.6 versus 79.3 and 74 versus 81.1), but these results were not found when compared to the French population (respectively 72.6 versus 69.8 and 74 versus 76.4). In the same way, mean scores for parent-proxy reports, emotional functioning at diagnosis were lower than in the healthy population when compared to US population (71.1 versus 81.3) while they were the same when compared to the French population (71.1 versus 71.8). This reflects the fact that, in the French reference population, the mean values for emotional, social and school functioning are generally lower than that of the US population reference (except for social functioning in parent-proxy reports), maybe due to cultural differences between the two populations and to the fact that the two populations are not exactly similar in term of age.

We studied HRQL scores observed on D1 and M6. As previously defined, these changes were expressed in term of MCID, which measures more closely the impact of differences observed in quality of life. In children's reports, we observed a statistically significant improvement in HRQL scores from D1 to M6 only for global and emotional scores (Δ respectively 4.5 for an expected MCID score 4.3 and 9.4 for an expected MCID score 8.3). No statistically significant differences were found in parent-proxy reports (Δ < MCID for all items). The percentages of patients that presented a PedsQL score perceived to be beneficial (score at M6 at least equal to the value of MCID) in terms of global, physical and psychosocial score were respectively 54.5%, 35.6% and 36.2%. In the same way, the proportion of patients that presented a PedsQL score perceived to be beneficial in term of emotional, social and school functioning were respectively 42.1%, 28.5% and 32.1%.

There was no minimal floor effect for ITP children. However, ceiling effects existed. Ceiling effects for healthy children highly satisfied with their quality of life were reported in the literature but with much greater proportions compared to ITP children at diagnosis 21 .

ICC between child and parent reports (Table 4) seemed better in M6 than at inclusion, with a better correlation in older children (more than 13 years old). For these teenagers, the main differences related to the physical score and the school functioning (upon inclusion) and the social and emotional functioning (in M6). The ICC between child self-report and parent proxy-report was measured as moderate to excellent.

ICC (intraclass correlation coefficients) are designated as < =0.4 poor to fair agreement, 0.41-0.60 moderate agreement, 0.61-0.80 good agreement, and 0.81-1.00 excellent agreement.

To determine their influence on HRQL scores and MCID, several parameters were analyzed: gender, age at diagnosis, preceding viral infection, initial ICIS score, initial place of treatment, and platelet count on M6. Change in quality of life was measured by MCID score (Table 5). Patients with an initial hospitalization had a lower score in social functioning (75% vs 25%, p = 0.05) and patients with a preceding viral infection perceived an improvement of quality of life (73.3% vs 26.7%, p = 0.05) in term of global score, at least equal to the MCID score at M6. In contrast, there was no statistically significant difference related to gender, age and ICIS score at diagnosis or initial treatment. Platelet count at M6 (< or >100.109/L) did not affect the MCID either.

ref = reference/β coefficients and SEs from logistic regression such that positive β coefficient means that predictor or increasing values of predictor are associated with better quality of life with MCID.