P value not match OR CI

I am running a multivariate LR and have categorical age variable with 6 categories (ref=0). My regression runs fine and everything seems to be in order except for this one variable. In the OR estimates the 95% confidence interval crosses 1 for category 3 vs 0; however, the p value for category 3 in the maximum likelihood estimate is 0.0013. Conversely, category 4 vs 0 has tight CI (0.30, 0.54) far from 1 but the p value is 0.21. A second variable with mutliple levels (yrscat) has one level that also doesn't have matching p and OR with the OR crossing 1 and the p value of 0.006.

Why are my OR CI not matching my p values?

My code is:
proc logistic data=...;
class sex (ref='0') agecat (ref='0') job2 (ref='0') hcw (ref='0') farmworker (ref='0') yrscat (ref='0');
model mex (ref='0') = sex agecat job2 hcw farmworker yrscat;
run;