In 2004, two systematic reviews drew attention to the outcomes of singleton pregnancies from assisted reproductive technology (ART) (Helmerhorst et al. BMJ 2004;328:261–5; Jackson et al. Obstet Gynecol 2004;103:551–63). Although their inclusion criteria differed, the conclusions were remarkably similar, demonstrating significantly worse pregnancy outcomes compared with non-assisted pregnancies. Both raised the question of how much of the poor outcome related to the characteristics of the women relying on ART and how much related to ART itself. This has remained a puzzle (Keirse & Helmerhorst. In: Pijnenborg et al., editors. Placental Bed Disorders. Cambridge: Cambridge University Press; 2010. p. 207–28) as there are several factors that hamper a clear-cut answer. Women who undergo ART differ in many respects from those who conceive naturally, not least in age, parity and the known or unknown reasons for their subfertility. ART involves different techniques, which may influence gametes, zygotes and implantation in several ways that are not fully understood. Not all singleton pregnancies after ART start as singletons. Some suffer from a vanishing twin syndrome that is associated with poorer outcomes. All is further compounded by an obstetric tendency to intervene in such pregnancies before problems arise (Keirse et al. Paediatr Perinat Epidemiol 2009;23:522–32). The study of DoPierala (BJOG 2016;123:1320–8), is a commendable attempt to disentangle the effect of subfertility from that of its treatment. Using a unique patient identification number, they linked women in a maternity database to women in a fertility centre database over nearly two decades. Linking exposures in one database to outcomes in another is always problematic. Although data on low birthweight and gestational age were missing for less than 0.02%, data on small for dates were missing for 4.6%. Only 44% of women attending the fertility clinic had received any treatment and, for 7.9%, it was unknown whether treatment had been instituted. In women without fertility centre history, there were 21 twin pregnancies per 1000 pregnancies [95% confidence interval (CI), 19.8–22.2] compared with only 14 per 1000 (95% CI, 13.I–15.1), including assisted conceptions, in Scotland in 2004 (EURO-PERISTAT project, 2008, www.europeristat.com). Nonetheless, the study convincingly demonstrates that a history of subfertility by itself is a risk factor for several adverse pregnancy outcomes, including pre-eclampsia, antepartum haemorrhage, placenta praevia, preterm birth, low birthweight and childbirth interventions. Although these data seem to indicate no difference in outcome between treated and non-treated subfertility, such a conclusion is not warranted in view of the concerns mentioned above. There is sufficient circumstantial evidence that ovarian stimulation is at least partially responsible for poor pregnancy outcomes after ART. Differences in outcome with natural conceptions seem to be similar for ovarian stimulation with and without in vitro fertilisation (IVF), whereas pregnancies from the transfer of thawed embryos, which is usually performed in natural cycles, appear to do better (Keirse & Helmerhorst, 2010). The Scottish data do not change the findings of a systematic review (Keirse & Helmerhorst, 2010) showing that different ART techniques have differential effects on the preterm birth rate. Although little can be concluded from the twin data, they reinforce the concept that the largest potential for reducing the frequency of adverse outcomes after ART is in the avoidance of multiple pregnancies, which – at least in theory – ought to be easier than with natural conceptions (Keirse & Helmerhorst, 2010). The International Committee of Medical Journal Editors (ICMJE) disclosure form is available online as supporting information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.