To evaluate the efficacy and safety of immunotherapy (hyposensitisation) in patients with severe summer hay fever. A randomised, double blind, placebo controlled study of a biologically standardised depot grass pollen extract. Allergy clinic, Royal Brompton and National Heart Hospital, London. 40 adults (mean age 35 years) with a history of severe grass pollen allergy uncontrolled by standard antiallergic drugs. Patients with perennial asthma were specifically excluded. Patients were randomised to receive either an active preparation (Alutard SQ, a grass pollen (Phleum pratense) extract) or placebo at a rate of two subcutaneous injections a week in increasing doses until a maintenance dose was reached. This maintenance dose was given once a month. Clinical efficacy was evaluated by symptom and drug diary cards, visual analogue scores during the grass pollen season, and a postseasonal assessment by the patients and a doctor. Conjunctival and skin sensitivity to local allergen provocation was measured before and after eight months of treatment. There was a highly significant decrease (median Alutard SQ v median placebo (95% confidence interval for difference between medians] in total symptom scores (p=0.001) in the Alutard SQ treated group (360 v 928 (238 to 825]. Significant differences were also found in total drug use (p=0.002, 129 v 627 (178 to 574]. Visual analogue symptom scores were also reduced in the active group (p=0.02, 2.2 v 5.5 (-4.8 to -0.5]. The postseasonal assessment, by either the doctor or the patients, showed a large improvement (p less than 0.001) in favour of Alutard SQ. Provocation tests showed a greater than 10-fold reduction for the active group in immediate conjunctival allergen sensitivity (p=0.001), a 40% decrease in early phase response (p=0.02), and a 57% decrease in the late phase (p=0.001) cutaneous response after intradermal allergen. A total of 523 active injections were given. There was one systemic reaction at 10 minutes after injection, which was rapidly reversed with intramuscular adrenaline. There was one mild delayed urticarial reaction at 2 1/2 hours. Immunotherapy is effective in patients with severe summer hay fever, but immediate anaphylactic reactions limit its use to specialised centres. Patient selection is extremely important, and chronic perennial asthma should be specifically excluded. As serious reactions occur within minutes a two hour wait for all patients after each injection seems unnecessary.