The aim of our study was to assess whether a non-invasive insulin injector could improve the metabolic control of ten diabetic children complaining of painful injections with syringe and needle. The cumulative study period amounted to 1347 days. Whereas a non-significant rise in insulin needs was observed (from 0.98 +/- 0.03 to 1.03 +/- 0.06 units/kg per day, mean +/- sem), mean HbA1c value remained unchanged (8.9% +/- 0.4% vs 9.0% +/- 0.5%). Jet injections were felt as less painful than those using syringe and needle (nine out of ten cases). This advantage was hampered by side-effects in eight out of ten cases such as episodes of glycoketonuria (six out of ten cases) leading to hospitalization in three patients. Other side-effects included inability to adjust injection pressure (four out of ten cases) and technical failure requiring an exchange of injector in five cases. The four children with most serious problems were significantly younger (P = 0.009) than other subjects. In conclusion, this type of injector should be discouraged in young diabetic children. For older children and adolescents, it may be an alternative to syringe and needle provided repeated detailed information and tight medical supervision is available.