RESULTS

Study 1: symptomatic ON

Characteristics of studied cohort

Among 4973 childhood leukemia survivors included in the L.E.A cohort on June 11, 2019, 129 (2.5%) had suffered from a symptomatic ON and all of them were included in study 1. Among patients with symptomatic ON, 66 had MRI performed for the diagnosis of ON. In 39 patients this MRI was available for central review at time of study n° 2. (Fig. 1).
The detailed characteristics of the 129 patients included in Study n°1 are described in Table 1. The median follow-up time for these patients was 9.97 years and that of the entire cohort was 10.94 years.
ON population consisted of mainly patients with ALL (88%) with a predominance of females (56.6%). Most patients were over 10 years old with a median age of occurrence of 13.3 years. One third of the patients received an allogeneic SCT prior to the onset of bone necrosis. Overall, 69% of patients had a multifocal involvement at ON diagnosis. The median time to onset of ON was 1.8 years after AL diagnosis. Among patients who received a SCT, necrosis was diagnosed on average 1.4 years after the transplantation.

Risk factors for osteonecrosis and determinants for multifocal character

We analyzed our cohort, after excluding the 365 patients treated for AML without SCT, since none of these patients had bone necrosis.
As a first step, we performed a comparative analysis of the 129 patients with ON with the 4479 patients of the L.E.A cohort without ON in univariate (Supplemental Table S1) and further in multivariate analysis (Fig. 2). SCT was associated with a higher risk of ON in univariate analysis (4.8% of ON for transplanted patients versus 2.3% for those not, p < 0.001), but no longer in multivariate analysis (OR 1.29, 95%CI [0.82-2.05], p=0.269). In univariate analysis, BMI at diagnosis (Z score) was also significantly associated with ON in the group of ALL without SCT (p=0.008) but no longer in multivariate analysis (OR 1.08, 95%CI [0.94-1.24], p=0.2). By contrast in multivariate analysis (Fig. 2), we showed that the diagnosis of AL after the age of 10 years was associated with a higher occurrence of ON (OR 22.46, 95%CI [13.8 -36.55] p <10-6). Females were also more often affected by ON (OR 1.8, 95%CI [1.23-2.58] p=0.002), but this predominance disappeared into the group of transplanted patients (OR 1.49, 95%CI [0.7-3.13], p=0.29). We finally highlighted that the presence of a relapse was associated with a more frequent occurrence of bone necrosis (OR 1.80, 95%CI [1.02-3.16], p=0.041).
In addition, we showed in the subgroup of patients who received a SCT (Fig.2B) a higher incidence of ON in patients over 10 years of age and who presented with chronic GVHD. We did not find any evidence of an increased risk of ON according to irradiation use.
In a second step, we looked for risk factors for multifocal involvement in our study population (Table 2). Multifocal involvement was not associated with a particular patient profile, but it was the most frequent presentation occurring in 89/129 (69%) patients suffering from ON.
We were also able to demonstrate that patients who presented ON were also those who were more likely to suffered from multiple sequelae (p<10-6) (Supplemental Table S2)

Quality of life

First, we compared the last assessment of QoL of patients with osteonecrosis with that of the general French population and of the L.E.A cohort, separately for adults, according to the SF-36 score (Fig. 3A), then for adolescents, according to the parents VSP-A score (Fig. 3B).
We obtained a QoL assessment for 88.7% of L.E.A patients (4088/4608), and for 118 of the 129 osteonecrosis patients. Data was collected from VSP-A parents score for 2087 patients of L.E.A, and 15 patients with ON, and from SF-36 for 1983 patients of L.E.A, and 103 patients with osteonecrosis. In comparison with the general adult population, ON led to a decrease in each parameter of the SF-36 score: physical (74.26 vs 94.76, p <10-6), social (73.14 vs 83.97, p<0.001), and emotional well-being scores (67.51 vs 87, p <10-6). Furthermore, this negative impact on QoL is also shown when comparing with patients of the L.E.A cohort, for most of parameters except mental health and emotional scores. (Fig. 3A)
These statements were also found when comparing adolescents with their healthy peers, mainly through a physical (physical well-being 44.16 vs 69.97, p=0.002) and friendship impact (48.93 vs 66.86, p=0.005) (Fig. 3B)
Secondly, we looked for the evolution of the QoL of patients with osteonecrosis during their lifetime. We therefore compared the results for the 37 patients for who two assessments of SF-36 questionnaire were available at two different times: first assessment after ON diagnosis with the last available assessment (Fig. 3C). The median time between these two evaluations was 45.6 months [11.6-127]. There was no difference in QoL score with time from ON diagnosis suggesting that the poor impact of ON on QoL occurred soon after ON and lasted.

Study 2: radiological description of joint damage at diagnosis

Pain was the main symptom identified in medical records as a diagnostic call point for ON, leading to the ordering of an initial standard radiography sometimes followed by an MRI.
Fifty-one patients were excluded for lack of MRI performed. Most of these suffered from osteonecrosis diagnosed before 2009 (40/51 = 72.5%), and for all before 2015. Twenty-seven MRI could not be read even though they had been performed, in connection with default of imaging storage since half of them were carried out before 2009, when a storage tool was generalized in French hospitals. Finally, radiological analysis could be performed for 39 patients, on 63 joints. Our radiological results mainly concerned patients whose leukemia was diagnosed after January 1, 2009: 32 patients out of 39 patients, which corresponded to 69% (32/46) of all patients diagnosed during this period. The characteristics of the 39 patients were not significantly different from those of the 90 patients who had no initial imaging of their necrosis, except for the delay between AL diagnosis and ON which was shorter for the patients included in the study n°2. (Supplemental Table S3).
For the 63 joints analyzed by MRI, we observed only 6 unconformities between both radiologists, which corresponded to a good inter-observer reproducibility with a kappa coefficient of 0.854. Radiological findings per joints and per patients are available in Supplementals Tables S4A and S4B.
The most often affected joints were the weight-bearing ones: knees and hips, with radiological involvement of more than one joint from the time of ON diagnosis in 56.4% of cases: 20 patients had bilateral involvement and 2 patients had multiple joint osteonecrosis (illustrated in supplemental fig.1).
Radiological severity was defined by the presence of V grade, i.e. joint deformity according to Niinimäki’s classification 16. Among the 14 patients with severe involvement, 8 had multifocal involvement. The hips injuries were more often severe (p =0.003) at the time of diagnosis of ON (Table 3). There was also a correlation between radiological severity and surgical management since patients with grade V disease, have more often benefited from joint replacement by prosthesis. (57.14% vs 16%, p = 0.012)
MRI follow-up was performed for 11 patients (28%), with a median delay of 10.9 months between the 2 imaging. There was a stage change for the same articulation on the follow-up MRI for only 2 patients, in whom stage IV worsened to stage V. Data are available in Supplemental Table S5.