Objective: To determine whether patients with multiple myeloma treated in three consecutive clinical trials of chemotherapy of the Finnish Leukaemia Group during 1979-85 had a more favourable prognosis than both patients treated in the trial area before the trials and those treated in the rest of Finland.

Design: Comparison of the time trend in survival of patients living in the trial area with that of patients living in the rest of Finland.

Setting: Trial area covered 17 of the 21 main hospital districts in Finland (serving more than two thirds of patients with multiple myeloma).

Patients: 663 Men and 690 women in the trial area and 318 men and 307 women in the reference area aged under 71 in whom multiple myeloma was diagnosed during 1959-85. In the trial area the disease was diagnosed in 455 men and 493 women in 1959-78 and in 208 men and 197 women in 1979-85; in the reference area it was diagnosed in 234 men and 227 women in 1959-78 and in 84 men and 80 women in 1979-85.

Main outcome measures: Five year cumulative relative survival rates during 1959-85, annual relative survival rates in the first seven years of follow up during 1979-85, and fitted annual relative survival rates in the first five years of follow up during 1959-85.

Results: During the first two years of follow up the annual relative survival rates did not differ between the two areas, but in the next five years of follow up patients in the trial area did better than those in the reference area. For cases diagnosed in 1979-85 the difference in the five year cumulative relative survival rates between patients in the trial area and those in the reference area was 10%, those in the trial area doing better. Generalised proportional hazards regression analysis of the first five years of follow up showed that the patients in the trial area had a survival advantage over those in the reference area. The model with the best fit included year of follow up, time of diagnosis, the joint effect of year of follow up and time of diagnosis, and the joint effect of area and time of diagnosis.

Conclusion: The patients in the trial area benefited from the clinical trials, which suggests that the use of a treatment protocol improves the end results of treatment. In other words, the results favour a systematic treatment schedule in preference to a schedule determined by the free choice of a clinician.