Prognostic and epidemiological factors in childhood leukemia : studies of p53 and MDM2 expression and of space-time clustering

Abstract

Although cure rates in acute childhood leukaemias have improved it remains a challenge to understand how molecular aberrations influence the prognosis of leukaemia and thus the choice of optimal therapy. Some studies have suggested that dysfunction of the tumour suppressor gene p53 or the oncogene MDM2 could indicate disease progression in children with leukaemia.

While these studies mainly have described uncommon pretranslational aberrations we examined p53 and MDM2 protein overexpression by immunohistochemical techniques in an unselected group of ALL children in the first study. As MDM2 overexpression was observed in high risk patients, in a second study we examined twenty-nine bone marrow samples in children with active, prognostically unfavourable and relapsed leukaemia. p53 protein was expressed in 12 patients and MDM2 in 17 patients. To put this in perspective, we investigated p53/MDM2 expression at diagnosis in 31 children who had maintained a long-term remission of ALL. Of these patients only 4 cases expressed p53, while 8 of 15 ALL-patients in relapse did (p=0.01). Four of these cases also expressed MDM2, while 10 of 15 relapsed ALLs did (p=0.0007). MDM2 and p53 expression was also seen in a higher proportion in the original diagnostic bone marrows of the relapse patients than in children with subsequent long-term remission. In conclusion these observations support the hypothesis that p53 and MDM2 are overexpressed in more aggressive leukaemias. Prospective studies are needed to investigate whether p53/MDM2 alterations at relapse represent secondary changes appearing during disease, or when present at diagnosis, if they are indicators of a poor outcome. Clustering of family cases of neoplasia could be a key to the genetic and environmental factors leading to cancer. By linking the Medical Birth Registry to the Cancer Registry in Sweden we conducted a nation-wide study including 2354 mothers whose children had had cancer, to assess whether these mothers had an increased risk of cancer compared with the general population. Our main findings revealed no general association between childhood cancer and cancer risks among mothers.

These registries were also used to study space-time clustering in childhood lymphatic leukaemias. If the etiological agent is infectious and the induction of leukaemia takes place perinatally, space-time clustering would be related to the time of birth rather than the time of diagnosis. We studied 645 recorded cases of childhood acute lymphoblastic leukaemia in Sweden 1973-1996, by using the close pair method of Knox. There was a significant excess of case-pairs close in date and place of birth in the 5-15-year age group. The group was expanded to 1020 recorded cases of childhood ALL and 293 cases of non-Hodgkin's lymphoma between 1973 -1996. A significant excess of case-pairs (p=0.01) was observed close in date and place of birth in the 4-14-year age group with ALL, but not when the cases of ALL and the non-Hodgkin's lymphomas were combined.

Although only a small proportion of the cases is involved in the clusters, in conjunction with other data these results indicate that childhood leukaemia can occur in clusters and strengthen the hypothesis that perinatal infections may be involved in a process leading to ALL.