Résumé : Objective: We compared official maternal mortality statistics with those from a special study covering all pregnancy-associated deaths in two European countries (Finland and France) and in two US states (Massachusetts and North Carolina) in 1999-2000 to characterize pregnancy-related deaths that are not included in official statistics. Study design: We linked the official ICD-10-based maternal mortality data for 84 deaths with study data on 404 pregnancy-associated deaths. Results: Of the pregnancy-associated deaths, 151 were pregnancy-related. We found 69 pregnancy-related deaths that had not been included as maternal deaths, and two deaths coded as maternal deaths that did not meet our definition for a pregnancy-related death. In total, 58 of these 69 deaths were from medical causes and 11 were from external causes or injuries (10 postpartum depression-related suicides and one accidental drug poisoning). The unreported deaths due to medical causes included 27 direct, 15 indirect, and two direct/indirect pregnancy-related deaths and 14 possibly pregnancy-related deaths. The most common causes of the unreported deaths due to medical causes were intracerebral hemorrhage (7 deaths), peripartum cardiomyopathy (4), pulmonary embolism (4) and pregnancy-induced hypertension (4). Conclusions: The collection of data on pregnancy-related and pregnancy-associated deaths is useful for countries with low maternal mortality figures. The use of various data-collection methods may substantially increase the quality of maternal mortality statistics. © 2006 Elsevier Ireland Ltd. All rights reserved.