The investigation of the association between breech presentation and cerebral palsy (CP) presents a major challenge because of the rarity of the outcome in a sub-population that only accounts for 3 to 4% of term deliveries. Randomized studies such as the Term Breech Trial1 are unlikely to ever recruit enough cases to gain sufficient power to examine such rare outcomes. Cohort studies remain our best source of data to establish associations, although they are unable to demonstrate causality. Using data from a national CP registry and the Medical Birth Registry of Norway, Andersen et al. tested the hypothesis that breech presentation is an independent risk factor for CP and whether this relates to mode of delivery. They also aimed to assess if breech presentation is associated with low Apgar scores, and severity and sub-type of CP.2 This study demonstrates a significant association between breech delivery and CP. No particular pattern of CP was associated with breech births. However, the study also demonstrates many of the difficulties inherent in the use of large data banks to test hypotheses. The first problem is that although there were initially 374 cases of CP diagnosed in the study population of over 170 000 births, only 245 were available with sufficient information to be included in the study. This degree of exclusion will always put significant limitations on the conclusions that can be drawn since it is not possible to know that missing data is not in some way a systematic error. Even in CP cases where information was considered sufficient for inclusion, information as basic as gestation (term or preterm) was missing in 3 out of 29 breech and 23 out of 216 vertex live births. Term breech infants had a relative risk of CP of 2.98 compared with term vertex infants. When considering all breech births the relative risk was just over 4 and this remained significant when controlled for gestational age, assisted conception, small for gestational age, and sex. The elements of this association remain unclear.3 To clarify how mode of delivery affects CP in breech presenting infants would require a larger sample size, comparisons including Caesarean section and vaginal birth in breech, breakdown by elective and emergency Caesarean section, and comparison of CP rates between breech and vertex infants delivered at term in the same way (e.g. emergency/elective Caesarean section term vertex compared with emergency/elective Caesarean section term breech). This would help to control for the factors that lead to choice of delivery mode. With so few cases of breech CP these breakdowns are difficult. Birth defects were nearly three times as common in infants presenting by the breech as in those presenting by the vertex. This association is well described. Antenatal diagnosis of a birth defect may well have an impact on planned mode of delivery. Knowledge antenatally, for instance, that an infant has a serious malformation is likely to influence whether a mother chooses a Caesarean section performed in the fetal interest. The strong association between birth defects and newborn encephalopathy,4 as well as CP,5 has the capacity to confound findings if adjustment for the presence of birth defects is not included in the analysis. This remains a difficulty with the Andersen et al. study. The authors have attempted to address the issue of CP alone as an outcome. When it comes to consideration of the impact of mode of delivery, CP alone is an inadequate endpoint since the risk to the fetus is of death or damage. Not only is death a crucial endpoint, but it is very possible that some of those who died might have gone on to develop CP should they have survived. The picture of outcomes is, therefore, incomplete and might lead to false assumptions about the usefulness or otherwise of Caesarean section in preventing CP among singletons presenting by the breech. Birth complications in breech presentation, e.g. cord prolapse or a trapped head, have severe consequences, including perinatal death, and need to be carefully considered along with CP in assessing the safety of a proposed mode of delivery. Consideration of CP alone may lead to a false impression of the safety of one mode over another. Following the Term Breech Trial very few vaginal breech deliveries are undertaken in developed countries. Between 1996 and 2005 the rate of vaginal delivery of breech presenting infants halved from over 20% to only 10%,6 most of whom were probably second twins. De-skilling of the obstetric workforce is a growing problem. A proportion of singletons who present by the breech will still either arrive at hospital too late to receive a Caesarean section or have mothers who choose not to have one. It could be argued that, soon, breech delivery will be beyond the experience of most obstetricians and midwives.