The basic principles of current idiopathic scoliosis treatment are three-dimensional correction and rigid fixation. Although it is accepted that Cotrel-Dubousset instrumentation (CDI) meets these goals, there is concern about the potential risk of trunk imbalance and spinal decompensation during the derotation manoeuvre. The results of 45 patients with idiopathic scoliosis treated with CDI between December 1988 and August 1992 were retrospectively analysed. Mean age was 14.3 years and mean follow-up period was 48.6 months. An average correction of 49.6% was achieved in the major curves. The best results were obtained in King type III curves, with a 69.4% correction. Spinal imbalance was evaluated by measuring lateral trunk shift (LT), shift of head (SH) and shift of stable vertebra (SS). Decompensation was measured by the increase in secondary curves. When all curve types were included, the average preoperative LT value of 1.96 vertebral units (VU) was brought down to 0.91 VU postoperatively, achieving a 55.9% correction. Fourteen patients had an SH value of zero preoperatively and remained balanced after instrumentation. Of the 41 remaining patients, 21 achieved an SH value of zero postoperatively. When all cases were included, the average preoperative SH value was 1.0 VU, which was corrected to 0.42 VU with CDI (69% correction). An average correction of SS of 75.5% was obtained, with the mean preoperative value of 0.73 VU being corrected to 0.19 VU. At the last follow-up visit, a secondary curve had formed above the major curve in one patient, and three patients had a junctional kyphosis. Loss of correction in the frontal plane correlated with loss of correction of LT. The rigid and semiflexible lumbar curves had a tendency to progress when they were not instrumented, especially in type II curves. Junctional kyphosis could be prevented when concave laminar claws were used in the thoraco-lumbar region. It was concluded that spinal decompensation and imbalance could be minimized with careful preoperative planning, avoidance of overcorrection and use of long instrumentation in double major curves.