Because of existence of oral temperature change and pH change and masticatory loading, it was unclear if thermocycling simulation only was sufficient to predict clinical microleakage of sealant. Objectives: The research was aimed to compare microleakage of sealant, which was performed clinically, with microleakage of sealant that was performed in vitro and subsequently subjected to thermocycling. Methods: Two experimental groups (n=25/each) were as follows; (i) clinical application of sealant (Concise®) on the occlusal surfaces of human premolars with 3-6 month function prior to extraction, and (ii) in vitro application of sealant on the extracted symmetrical premolars, followed by 1,000 cycles of thermocycling with 5 and 55°C with 30 sec dwelling time. Then, all specimens were stained in 2% basic fuchsin for 24 h, and sectioned vertically (bucco-lingual) into 3 pieces, yielding 4 surfaces/specimen. Microleakage was assessed under 25x stereomicroscope as follows: 0 (no microleakage), 1 (questionable microleakage), 2 (partial microleakage), 3 (full microleakage). Wilcoxon match-paired statistics was used to determine differences of microleakage of clinical and in vitro sealant, α=0.05. Results: The results showed that (i) microleakage of clinical sealant (average microleakage=1.50, SE=0.16) was greater than that in vitro (average microleakage=1.03, SE=0.16) but not statistically significant, and (ii) 63 surfaces of the total 91 surfaces of clinical sealant (69.2%) and 46 surfaces of the total 92 surfaces of in vitro sealant (50%) showed microleakage. Conclusions: It can be concluded that incorporation of only thermocycling for testing of microleakage of in vitro sealant may slightly underestimate microleakage of short-term clinical sealant but perhaps is an acceptable method to determine microleakage of sealant. (Financially supported by Prince of Songkla University, Songkla, Thailand)