AIM: To determine the value of the polymerase chain reaction (PCR) technique and the analysis of intraocularly produced antibodies by calculating a Goldmann-Witmer quotient (GWq) as diagnostic assays in the confirmation of a clinically diagnosed cytomegalovirus (CMV) retinitis in a group of unselected AIDS patients. METHODS: Eleven samples of undiluted ocular fluid, obtained from nine AIDS patients with a clinically diagnosed CMV retinitis were analysed for the presence of genomic DNA from CMV, HSV-1, VZV, and EBV by PCR. Nine of these samples were analysed for the presence of locally produced IgG antibodies against these herpesviruses by calculating a GWq. Ten samples obtained from patients with various entities of clinical non-herpetic uveitis and 17 samples of aqueous humour obtained at cataract surgery were used as controls. RESULTS: In 10 out of 11 samples from AIDS patients (91%) the presence of CMV DNA was demonstrated. In four out of nine (44%) patients this was accompanied by CMV DNA in the blood indicating a CMV viraemia. In one sample, VZV DNA was detected and in another sample both CMV and VZV DNA were detected. No HSV-1 or EBV DNA could be demonstrated in these 11 samples. In contrast, simultaneous analysis of locally produced IgG antibodies against herpesviruses could not confirm the initial diagnosis of CMV retinitis. Ocular fluid samples obtained from 10 control uveitis patients were negative for DNA from CMV, VZV, and EBV by PCR. In one of 10 uveitis control samples HSV-1 DNA was detected; antibody analysis did not confirm this. In the uveitis control group, a significant GWq was calculated in one sample for HSV-1 and in another sample for VZV. The cataract control samples were all herpesvirus DNA negative by PCR. CONCLUSIONS: To establish the diagnosis of CMV retinitis in AIDS patients, ophthalmoscopic examination is a sensitive method. In confirming a diagnosis in indistinctive cases, application of a PCR assay detecting CMV DNA is a more sensitive method than analysis of locally produced antibodies by calculating a GWq. In clinical non-herpetic uveitis, secondary release of HSV-1 and VZV should be considered requiring additional therapeutic anticipation.