An attempt has been made to explore selectivity requirements for cyclooxygenase-2 (COX-2) versus cyclooxygenase-1 (COX-1) binding of terphenyl methyl sulfones and sulfonamides using electrotopological state (E-state) index and suitable indicator parameters. Multiple linear regression analyses produced statistically acceptable equations: the best relation based on 'all-possible-subsets regression' for COX-1 binding (n=18) showed predicted variance and explained variance of 0.675 and 0.777, respectively, while in case of the best equation for COX-2 binding (n=38), these values rose to 0.842 and 0.874, respectively. For the selectivity relation (n=17), predicted variance and explained variance values were 0.601 and 0.687, respectively. Based on the results of the analyses, three important sites have been suggested: sites A (methylsulfonyl or aminosulfonyl moiety), B (central phenyl ring), and C (terminal phenyl ring containing different substituents). All three sites are important for COX-2 binding while sites B and C are important for COX-1 binding. For COX-2 selectivity, only site C plays an important role. The study shows the utility of E-state index in developing statistically acceptable model having direct physicochemical significance.