We conducted a double-blind, randomized phase II study to evaluate the safety and activity of combination therapy with N-butyl-deoxynojirimycin (SC-48334) (an alpha-glucosidase I inhibitor) and zidovudine versus zidovudine alone. Patients with 200 to 500 CD4 cells/mm3 who tolerated < or = 12 weeks of prior zidovudine therapy received SC-48334 (1000 mg every 8 h) and zidovudine (100 mg every 8 h) or zidovudine and placebo. Sixty patients received combination therapy and 58, zidovudine and placebo. Twenty-three patients (38%) and 15 (26%), in the combination and zidovudine groups, respectively, discontinued therapy (p = 0.15). The mean SC-48334 steady-state trough level (4.04 +/- 0.99 micrograms/ml) was below the in vitro inhibitory concentration for human immunodeficiency virus (HIV). The mean increase in CD4 cells at week 4 was 73.8 cells/mm3 and 52.4 cells/mm3 for the combination and zidovudine groups, respectively (p > 0.36). For patients with prior zidovudine therapy, the mean change in CD4 cells in the combination and zidovudine groups was 63.7 cells/mm3 and 4.9 cells/mm3 at week 8 and 6.8 cells/mm3 and -45.1 cells/mm3 at week 16, respectively. The number of patients with suppression of HIV p24 antigenemia in the combination and zidovudine groups was six (40%) and two (11%) at week 4 (p = 0.10) and five (45%) and two (14%) at week 24 (p = 0.08), respectively. Diarrhea, flatulence, abdominal pain, and weight loss were common for combination recipients.(ABSTRACT TRUNCATED AT 250 WORDS)