AIM To evaluate the possibility of detecting Helicobacter pylori (Hp) in human saliva by PCR, to observe the relationship between the positive rate of Hp in gastric mucosa and the grades of gastritis, and to compare the positive rate of Hp by PCR with other methods. METHODS Biopsy specimens of 122 patients (53 duodenal ulcer, 21 gastric ulcer, 13 gastroduodenal ulcer and 35 chronic gastritis) from gastric antrum and body were studied with PCR, rapid urease test and Giemsa stain through gastric endoscopy to detect Hp. The saliva of patients was only detected for Hp with PCR. The patients were asked not to take antibiotics, bismuth and H + K + pump inhibitor in 2 weeks before endoscopy. The positive rates of Hp were compared with U test among different methods, and the relationship of grades of gastritis and extents of Hp infection, with χ 2 test. RESULTS Specimens from 105 patients were detected for Hp with PCR. The positive rate of Hp was 5 7% in the saliva and 61 0% in gastric specimens, which was significantly higher than that with Giemsa stain (49 2%, P 0 05) and lower than that with rapid urease test (75 0%, P 0 05). On the basis of positive Hp with rapid urease test and Giemsa stain at the same time, 6 of 51 (11 8%) were positive for Hp in the saliva and 49 of 56 (87 5%) in the gastric mucosa, meanwhile, the false positive rate was zero in saliva and 22 2% in gastric mucosa. The positive rate of Hp with PCR in gastric mucosa was 83 3% in severe gastritis, 69 8% in moderate gastritis, and 42 2% in mild gastritis ( P 0 05 among three kinds of gastritis). The Hp positive rate with Giemsa stain in gastric mucosa was strongly proportional to grades of the gastritis ( P 0 01). In gastric diseases, the positive rates of Hp with PCR and Giemsa stain were the highest in duodenal ulcer, and in gsatric ulcer, atrophic gastritis and superficial gastritis in turn. CONCLUSION Hp exists in saliva. PCR detection may be an effective, non invasive method for direct diagnosis of Hp infection and is better than other methods in sensitivity and specificity.