Optimization of a previously reported thrombin inhibitor, 9-hydroxy-9-fluorenylcarbonyl-l-prolyl-trans-4-aminocyclohexylmethylamide (1), by replacing the aminocyclohexyl P1 group provided a new lead structure, 9-hydroxy-9-fluorenylcarbonyl-l-prolyl-2-aminomethyl-5-chlorobenzylamide (2), with improved potency (K(i) = 0.49 nM for human thrombin, 2x APTT = 0.37 microM in human plasma) and pharmacokinetic properties (F = 39%, iv T(1/2) = 13 h in dogs). An effective strategy for reducing plasma protein binding of 2 and improving efficacy in an in vivo thrombosis model in rats was to replace the lipophilic fluorenyl group in P3 with an azafluorenyl group. Systematic investigation of all possible azafluorenyl P3 isomers and azafluorenyl-N-oxide analogues of 2 led to the identification of an optimal compound, 3-aza-9-hydroxyfluoren-9(R)-ylcarbonyl-l-prolyl-2-aminomethyl-5-chlorobenzylamide (19b), with high potency (K(i) = 0.40 nM, 2x APTT = 0.18 microM), excellent pharmacokinetic properties (F = 55%, T(1/2) = 14 h in dogs), and complete efficacy in the in vivo thrombosis model in rats (inhibition of FeCl(3)-induced vessel occlusions in six of six rats receiving an intravenous infusion of 10 microg/kg/min of 19b). The stereochemistry of the azafluorenyl group in 19b was determined by X-ray crystallographic analysis of its N-oxide derivative (23b) bound in the active site of human thrombin.