r is basically tested for significance with null hypotesis that ro is 0. Then statistics:
r*sqrt(n-2)/(1-r^2)
is calculated
and if this statistics fall into (-inf,-q) or (q,+inf) then null hypotesis shall be rejected
the 1-(alfa/2) quantile q of t-Student distribution of n-2 degrees of freedom could be found from from excel function.
as you need 99.73% level of confidence (very high indeed - have you considered a smaller one, because with so high confidence level you are quite prone to second type errors) so alfa = 0.27% you will get the quantile q from:
=TINV(0.27%,n-2)
(of course instead of n-2 write your number of XY pairs decreased by 2)