Summary The 'true' incidence of spontaneous regression of neuroblastomas is uncertain. However, the frequency of spontaneous regression is important when the benefits of screening procedures are considered. In the population-based Danish neuroblastoma survey 1943-80, spontaneous regression was documented in less than 2% of cases. However, the 'true' incidence may be higher. The epidemiological findings of increased incidence and survival rates with an unchanged mortality rate may suggest the inclusion of borderline lesions among 'truly' malignant neuroblastomas in recent decades in Denmark. However, it is more likely to be a result of improved diagnosis, changes in the social composition of the population and possibly unidentified environmental agents. However, if some premalignant lesions in fact had been included, they are most likely to be stages I-II tumours of infancy. In this study we describe cases of spontaneous regression of neuroblastoma from the Danish population-based survey 1943-80. Numerous anecdotal reports of cases of spontaneous remission of neuroblastomas have been published. For example, a Danish patient diagnosed in 1941 has been the subject of several reports (Hansen, 1953;Rosendal, 1942;Visfeldt, 1963). Neuroblastomas undergoing spontaneous regression form 17% of the cases of spontaneous remission in man collected from the literature by Everson and Cole (1966). In a review of cases registered in Children's Cancer Study Group, Evans et al. (1976) estimated the frequency to be 8% of cases (including stage IV-S cases). Pritchard and Kemshead (1983), excluding stage IV-S cases, found the frequency to be only 1-2% in their experience. However, a population-based study has not been carried out so far.Most childhood neuroblastomas are likely to be congenital (Birch et al., 1980;Carlsen, 1988a;Carlsen et al., 1986a;Rubin, 1968;Sutow, 1958;Wilson & Draper, 1974), and the prognosis is poor for most patients over the age of 2 years. However, the benefits of a mass screening in infants (Sawada et al., 1984a(Sawada et al., , 1987aScriver et al., 1987;Woods & Tuchman, 1987) has been questioned because it is argued that some cases detected at screening might have subsequently regressed spontaneously (Norman et al., 1987;Pastore et al., 1984).The Danish population-based survey 1943 -80 (Carlsen, 1986, 1988bCarlsen et al., 1986a Carlsen et al., , b, 1987) offers a unique opportunity to estimate the frequency of spontaneous regression from clinical overt disease in an unselected population of neuroblastomas of childhood. This study reports and discusses cases of confirmed regression and also the questionable cases. The study also considers age, stage and fate of cases found incidentally by abdominal examination or chest X-rays, as these cases are suggested to be equivalent to the cases found by screening (Kosloske et al., 1987;Sawada et al., 1988). As data from the Danish survey indicate a zerotime shift (Bailar & Smith, 1986;Feinstein et al., 1985) in the study period (Carlsen, 1986;Carlsen et al., 1...