Aims: To examine current BMI and various aspects of BMI history as pre-screening tools for undiagnosed diabetes in Japanese individuals. Methods: This cross-sectional study included 16 226 men and 7026 women aged 30-75 years without a self-reported history of clinician-diagnosed diabetes. We estimated the probability of having undiagnosed diabetes (fasting glucose ≥ 7.0 mmol/l and/or HbA&lt
inf&gt
1c&lt
/inf&gt
≥ 48 mmol/mol (≥ 6.5%) for the following variables: current BMI, BMI in the early 20s (BMI&lt
inf&gt
20y&lt
/inf&gt
), lifetime maximum BMI (BMI&lt
inf&gt
max&lt
/inf&gt
), change between BMI in the early 20s and current BMI (ΔBMI&lt
inf&gt
20y-cur&lt
/inf&gt
), change between BMI in the early 20s and maximum BMI (ΔBMI&lt
inf&gt
20y-max&lt
/inf&gt
), and change between lifetime maximum and current BMI (ΔBMI&lt
inf&gt
max-cur&lt
/inf&gt
). Results: The prevalence of undiagnosed diabetes was 3.3% (771/23252) among participants. BMI&lt
inf&gt
max&lt
/inf&gt
, ΔBMI&lt
inf&gt
20y-max&lt
/inf&gt
and current BMI (1-sd increments) were more strongly associated with diabetes than the other factors (multivariate odds ratio 1.58 [95% CI 1.47-1.70] in men and 1.65 [95% CI 1.43-1.90] in women for BMI&lt
inf&gt
max&lt
/inf&gt

multivariate odds ratio 1.47 [95% CI 1.37-1.58] in men and 1.61 [95% CI 1.41-1.84] in women for ΔBMI&lt
inf&gt
20y-max&lt
/inf&gt

multivariate odds ratio 1.47 [95% CI 1.36-1.58] in men and 1.63 [95% CI 1.40-1.89] in women for current BMI). The probability of having diabetes was markedly higher in those with both the highest tertile of BMI&lt
inf&gt
max&lt
/inf&gt
and greatest ΔBMI&lt
inf&gt
20y-max&lt
/inf&gt

however, a substantially lower likelihood of diabetes was observed among individuals with the lowest and middle tertiles of current BMI (&lt
24.62 kg/m&lt
sup&gt
2&lt
/sup&gt
in men and &lt
22.54 kg/m&lt
sup&gt
2&lt
/sup&gt
in women). Conclusions: Lifetime maximum BMI and BMI changes from early adulthood were strongly associated with undiagnosed diabetes. Adding BMI history to people's current BMI would improve the identification of individuals with a markedly higher probability of having undiagnosed diabetes.