OBJECTIVES: There have been several attempts to classify cause of death (CoD) in stillbirth; however, all such systems are subjective, allowing for observer bias and making comparisons between systems challenging. This study aimed to examine factors relating to determination of CoD using a large dataset from two specialist centers in which observer bias had been reduced by classifying findings objectively and assigning CoD based on predetermined criteria. METHODS: Detailed autopsy reports from intrauterine deaths in the second and third trimesters during 2005-2013 were reviewed and findings entered into a specially designed database, in which CoD was assigned using predefined objective criteria. Data regarding CoD categories and factors affecting determination of CoD were examined. RESULTS: There were 1064 intrauterine deaths, including 246 early intrauterine fetal deaths (IUFD) (< 20 weeks), 179 late IUFDs (20-23 weeks) and 639 stillbirths (≥ 24 weeks' gestation). Overall, around 40% (n = 412) had a clear CoD identified, whilst around 60% (n = 652) were classified as 'unexplained', including around half with identified risk factors or lesions of uncertain significance, with the remaining half (n = 292 (45%)) being entirely unexplained. A stepwise increase in the proportion of unexplained deaths was observed with increasing maceration. Black and Asian women had significantly greater proportions of deaths due to ascending infection, whilst women aged over 40 years had significantly increased placenta-related CoDs. There was no significant difference in CoD distribution according to maternal body mass index or with increasing postmortem interval. Around half of those with an identifiable CoD could be identified from clinical review and external fetal examination or imaging, with most of the remainder being determined following placental examination. CONCLUSIONS: Based on objective criteria, many intrauterine deaths throughout gestation remain unexplained despite autopsy examination. The rate of unexplained death varies from around 30% to 60% depending on interpretation of the significance of features. CoD determination is dependent on both the classification system used and subjective interpretation, such that variation in the proportion of 'unexplained' cases is based largely on speculation regarding mechanisms of death. Novel methods to determine objectively the mechanism of death at postmortem examination are required. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.