OBJECTIVE--To determine the factors associated with the occurrence of Pneumocystis carinii pneumonia (PCP) in Wegener's granulomatosis (WG). METHODS--We retrospectively compared a group of 12 patients with WG and PCP (PCP group), with 32 WG patients without PCP followed over the same period in the same centres (control group). RESULTS--The mean delay of onset of PCP after the start of the immunosuppressive therapy was 127 (SD 128) days. Before treatment, the clinical and biological features of the two groups were similar, except for the mean lymphocyte count which was lower in the PCP group than in the control group (1060/mm3 v 1426/mm3; p = 0.04). During treatment, both groups were lymphopenic. There was a significant difference between the lowest absolute lymphocyte count in each group (244/mm3 in the PCP group v 738/mm3 in the control group; p = 0.001). During the first three months of treatment, the lymphocyte count was less than 600/mm3 at least once in 10 of the 12 patients in the PCP group and in 11 of the 32 patients in the control group (p < 0.01). The mean cumulative dose of cyclophosphamide was greater in the PCP group than in the control group at the end of both the second (1.55 mg/kg/day v 0.99 mg/kg/day; p = 0.05) and the third (1.67 mg/kg/day v 0.97 mg/kg/day; p = 0.03) months. However, in multivariate analysis, the only two factors independently and significantly associated with the occurrence of PCP were the pretreatment lymphocyte count (p = 0.018) and the lymphocyte count three months after the start of the immunosuppressive treatment (p = 0.014). CONCLUSIONS--The severity of lymphocytopenia before and during immunosuppressive treatment is the factor best associated with PCP in WG.