Osteosarcoma is exceedingly rare in young children. In France, Desandes et al found an incidence of 9.2 cases per million adolescents , compared to only 0.4 per million under-fives . In our small series of 15 patients the sex ratio was 1.1 in favor of males, compared to 2.1 in adolescents in France . The tumour arose in the diaphysis in 6 cases, compared to only 4% of cases in Pakos' series of 2680 patients .
Factors predisposing to cancer are a question of interest in patient developing an osteosarcoma at young age. No genetic predisposition was found in our population, but germinal mutation of Rb and p53 have only been investigated in one child. Another 3-year-old child had constitutional tall height (+4 SD). Osteosarcoma usually occurs during periods of rapid growth, but the relationship with height remains controversial, recent retrospective studies having given contradictory results. Buckley found no consistent relation with height in a case-control study of 152 children with osteosarcoma , while Gelberg found a significant positive association with height one year before diagnosis . Adult patients with acromegaly have been reported to be subject to osteosarcoma . In our series, one child was treated with growth hormone (GH) for 3 years before diagnosis, following intrauterine growth retardation and a polymalformative syndrome, without biological GH deficiency. Due to this syndrome, a genetic predisposition to tumour cannot be excluded. However isolated cases of osteosarcoma have been reported during GH treatment, but no formal relationship has been established [26, 27]. Carel et al reported that GH-treated children had a relative risk of 13.8 for bone cancer . The benefit of GH treatment probably outweighs the risks, but the indication should be examined closely in case of personal or familial risk factors.
The average interval between symptom onset and diagnosis was 2 months in our patients, in keeping with other studies [13, 29], but longer than in Kager et al report . Two children had mistakenly been treated for benign tumours, by infiltration or resection. Although benign tumours are more frequent than neoplasms in young children , the possibility of malignancy must be kept in mind.
Metastasis was found at diagnosis in 40% of the children in this study, a rate higher than reported in older children (10 to 20%) , or in young children by other groups [15, 16].
The osteoblastic histologic type predominated in our series (73%), as in older patients and in the recent german study in young children . Other studies found fibroblastic type  or telangiectasic type  as predominant subtype. However, in those studies, centralized slide review was not performed.
As this study spans a lengthy period, treatment was heterogeneous. However, all the children who received chemotherapy (14 out of 15) were treated with drugs known to be effective in osteosarcoma (HDMTX, ifosfamide, cisplatin or doxorubicin). A good tumour response to pre-operative chemotherapy was obtained in only 36% of our patients, while similar regimens have been reported to give good responses in 56% to 64% of patients with non-metastatic osteosarcoma [19, 21, 32] and in 42% of patients with metastatic forms . This would suggest that preoperative chemotherapy is less effective in younger children. Likewise, Cho  reported good responses in only 2 (20%) of 10 children under 7 years of age. These results, together with the high frequency of metastases at diagnosis promote the hypothesis of a different tumour biology in young children, with more aggressive disease. An other explanation could be differences in chemotherapy metabolism in young patients. It has been demonstrated that systemic methotrexare and doxorubicin clearance tended to be lower in very young children , and quite recently, Crews et al found that in children and young adults with osteosarcoma, a lower methotrexate clearance was associated with lower probability of survival . However recent studies did not find any difference in histological response in young patients compared to older patients [14, 16] and previous methotrexate pharmacokinetics analyses stated that high peak level were associated with better outcome .
No major acute adverse effects of chemotherapy were noted, but a one-year-old boy developed late neurotoxicity. His peripheral neuropathy was attributable to ifosfamide, which is known to provoke painful peripheral sensory neuropathy , while his late convulsive encephalopathy and mental regression were attributable to HDMTX. Severe leukoencephalopathy has been described with this drug and cannot be avoided by folinic acid supplementation .
As reported by others in young children, we found a high amputation rate in our study, explained by tumour progression in some cases (n = 5), but also by the complexity of surgical reconstruction in young patients, and by the lengthy study period. Indeed, the frequency of amputation appears to be decreasing over time: it was 51% in a French study in 1988  and only 6% in the Os 94 trial .
Owing to the small number of patients and the lack of quality-of-life assessment, the functional outcome of children who had lower-limb surgery was difficult to assess. The 3 children who had limb-sparing surgery had a high rate of mechanical complications. Skeletal maturity is an important determinant of functional outcome in children with greater growth potential. Limb-sparing procedures are more problematic than in adults, and remain a challenge in very young children [13–16, 32, 39].
The overall and event-free survival rates in this study are difficult to interpret, given the small number of patients, but our results seem similar to those reported by Kager et al . It is noteworthy that 7 children (45%) died of their disease. Currently, the reported survival rate among children treated for osteosarcoma is about 19% for metastatic patients  and 76% at 5 years for non metastatic patients . In Germany, Bielack et al reported an overall survival rate of 65% at 5 years . In the US study by Mirabello et al, the overall survival rate was 61% at 5 years between 1973 and 2004 among patients under 24 years of age . In very young children, some authors did not find difference in survival compared to older patients [12–14, 16]. However, in a registries based study, Worch et al reported that in non metastatic patients, 5-year overall survival estimate was 51.9% for children who were 5 years of age or younger at diagnosis versus 67.3% for patients ages 6-19 years . Metastasis at diagnosis is a well-known factor of poor prognosis. In our series only 3 out of 6 metastatic patients were alive at the cut-off date for this analysis. The response to neoadjuvant chemotherapy was also found to have a prognostic influence in most studies. In our study, 1 of the 4 patients who had a good response died, compared to 4 of the 7 patients with a poor response. Kager et al also found a negative impact of poor histological response on survival . Six (50%) of the 12 children who obtained full remission in our series relapsed. This is a higher rate than in the Hartford's series of young children treated in the United States (25%) , but the same observation was done in Kager's study, who reported 11 recurrences among 23 patients (48%) who achieved complete remission . Our results confirm that pattern of recurrence seems to be similar to older children with lung and bone involvement.