Résumé : OBJECTIVE: Spinal cord arteriovenous shunts (SCAVSs) are currently classified according to their morphological features. Certain shunts cannot be fully integrated into the predetermined categories that are usually described, however. Can these classifications be reevaluated on the basis of recent anatomic, biological, and genetic advances? METHODS: We reviewed the clinical and radiological files for 155 SCAVSs that were treated at Hôpital Bicêtre between 1981 and 1999. The lesions were examined with respect to their number (single or multiple), their primary architectural type (nidus or fistula), and their possible links with associated metameric lesions. RESULTS: All SCAVSs were either arteriovenous malformations or fistulae, with the latter being either micro- or macrofistulae. All SCAVSs corresponded to three categories, i.e., genetic hereditary lesions (macrofistulae and hereditary hemorrhagic telangiectasia), genetic nonhereditary lesions (all of which were multiple lesions with metameric or myelomeric associations), and single lesions (which could represent incomplete presentations of one of the previous groups). Of the SCAVSs in our series, 81% were single lesions and 19% were multiple; among these, 59% were true intradural shunts with metameric features. Ten cases of Cobb syndrome, three cases of Klippel-Trenaunay syndrome, and two cases of Parkes-Weber syndrome, all with associated cord lesions, were observed. Nineteen percent of SCAVSs were fistulae; 23% of those were macrofistulae, of which 83% were related to Rendu-Osler-Weber disease. CONCLUSION: It seems legitimate to propose a categorization that takes into consideration a primary malformation (nidus or fistula) that evolves with time and in which angioarchitectural changes occur. Recognition of the factors originally responsible for the shunt (e.g., genetic hereditary or genetic nonhereditary) allows a different classification of SCAVSs.