It is therefore difficult to separate the effects of NSAIDs from the biological impacts resulting from AS. Moreover, it remains controversial whether NSAIDs is associated with an increased risk of stroke. A recent large-scale meta-analysis of 280 trials of NSAIDs versus placebo and 474 trials of one NSAID versus another NSAID showed there was no evidence that any NSAID, including selective COX-2 inhibitors and traditional NSAIDs, significantly increased the risk of stroke. Therefore, the use of NSAID was not included in our analysis. In the present population-based study, the estimated prevalence of AS was 0.12% using case definition that requires at least two ambulatory visits with a principal diagnosis of AS in 2001. This prevalence estimate is relatively lower than that obtained from a community-based survey on the prevalence of rheumatic diseases in Taiwan which used a 2-stage screening process in 1992. In that study, the estimated prevalence of AS in the adult Taiwanese population ranged from 0.19 to 0.54%. Since some AS patients with mild symptoms who did not seek medical service would not be recorded in the NHI database, our study may tend to recruit patients with more severe or active AS, and it can be expected that the estimated prevalence of AS in our study would be lower than that reported from the previous community-based survey. The strength of the present study was the use of a longitudinal population-based NHI database. The NHI program is a singlepayer compulsory social insurance program with considerably high coverage rate in Taiwan. The barrier to medical access is negligible because the NHI system allows patients to visit any clinic or hospital freely without referral by a general practitioner, and patients pay only about $5�C$15 USD at each visit. Atractylenolide-III Considering the neurological deficit and functional disability related to stroke, and the minimal barrier to medical access in Taiwan, it can be expected that most patients who 
developed stroke would seek medical help and would be captured in the NHI database, which enabled us to identify all incident cases of stroke and establish a temporal relationship between AS and ischemic stroke. Nevertheless, several limitations should be acknowledged. First, the diagnoses of AS, ischemic stroke, and medical comorbidities were determined using the ICD codes from the NHI claim database, and there may be concern about the diagnostic accuracy of the database. However, the Bureau of NHI has formed different audit commi ees that make it a rule to randomly sample the claims data from every hospital and to review charts on a regular basis to verify the diagnostic validity and quality of care. In addition, one validation study that evaluated the validity of the NHI database for patients with a principal diagnosis of ischemic stroke showed that the NHI database appears to be a valid resource for population-based research in ischemic stroke. Accordingly, the NHI claim database is an established research database and independent studies have demonstrated the validity of the data. Furthermore, we Benzoylpaeoniflorin performed sensitivity analyses using various case definitions, and found that AS was consistently linked to an increased risk of developing ischemic stroke.