We examined the association between prostate-specific antigen (PSA) and daily mean ambient temperature on the day of the test in healthy men who had three annual checkups. We investigated 9,694 men who visited a hospital for routine health checkups in 2007, 2008, and 2009. Although the means and medians of ambient temperature for the three years were similar, the mode in 2008 (15.8 °C) was very different from those in 2007 and 2009 (22.4 °C and 23.2 °C). After controlling for age, body mass index, and hematocrit, a multiple regression analysis revealed a U-shaped relationship between ambient temperature and PSA in 2007 and 2009 (P<0.001 and P=0.004, respectively), but not in 2008 (P=0.779). In 2007, PSA was 13.5% higher at 5 °C and 10.0% higher at 30 °C than that at 18.4 °C (nadir). In 2009, PSA was 7.3% higher at 5 °C and 6.8% at 30 °C compared with the level at 17.7 °C (nadir). In logistic regression analysis, a U-shaped relationship was found for the prevalence of a higher PSA (>2.5 ng/mL) by ambient temperature, with the lowest likelihood of having a high PSA at 17.8 °C in 2007 (P=0.038) and 15.5 °C in 2009 (P=0.033). When tested at 30 °C, there was a 57% excess risk of having a high PSA in 2007 and a 61% higher risk in 2009 compared with those at each nadir temperature. We found a U-shaped relationship between PSA and ambient temperature with the lowest level of PSA at 15-20 °C.