The evidence base for miscarriage diagnosis: better late than never

On average it takes 10–20 years for knowledge generated by trials to be incorporated into clinical practice1. There is a real risk that, by the time clinical practice evolves to be 'evidence-based', the evidence itself may have changed. This issue of the Journal includes a series of four papers on the topic of miscarriage diagnosis—the commonest emergency indication for an ultrasound in pregnancy. I have little hesitation in suggesting that these papers may well be the most practice-changing that UOG has published. In 1995 Stuart Campbell, the founding editor of this Journal, gave evidence to a public enquiry into incorrectly diagnosed miscarriages in the UK2. The authors suggested the following guideline for non-viability: a mean sac diameter (MSD) > 20 mm with no evidence of an embryo/yolk sac or an embryo of CRL > 10 mm with no evidence of heart pulsations, both on two separate occasions at least 7 days apart. Campbell and colleagues were careful not to be prescriptive; they did not state that these criteria were diagnostic, but rather 'highly suggestive' of pregnancy failure. The Royal College of Obstetricians and Gynaecologists' 2006 guideline uses an empty sac with MSD of ≥ 20 mm or lack of cardiac activity with an embryo ≥ 6 mm as a miscarriage3. The American College of Radiologists' 2009 guideline uses an empty sac with MSD > 16 mm or an embryo with a CRL > 5 mm without cardiac activity as a miscarriage4. The levels of evidence cited to support these cut-offs are at best weak to moderate, which would explain the variation in guidelines. Would it not be worrying if the criteria to pronounce death at any other stage of human life were as variable as those that exist currently in early pregnancy? The first article in the series presented in this Journal is a systematic review of the available literature on the ultrasound diagnosis of miscarriage. Jeve and colleagues demonstrate the lack of high-quality, prospective data on which to base guidelines for the accurate diagnosis of early pregnancy demise5. Not only are there only eight small studies, they are some 20 years old and use unreliable reference standards for the confirmation of pregnancy loss. The authors correctly conclude that clinicians cannot be certain that the current guidelines afford them a 100% specificity in the diagnosis of miscarriage. That is to say that it is possible that some women with a viable pregnancy will be inappropriately labeled as having a miscarriage. In the first of a pair of related papers, Abdallah and colleagues assessed over 1000 women and demonstrated that using an empty sac with MSD > 20 mm, there would be one viable pregnancy in every 200 cases classified as a miscarriage6. Worryingly, the equivalent false-positive rate for an empty sac with MSD > 16 mm was 4.4%. The authors suggest that, by slightly increasing the MSD and CRL sizes to define a miscarriage, the false-positive rates fall to zero. However, because of the small number of patients studied at the suggested cut-offs, the 95% confidence intervals for specificity still remain wide. Most guidelines stipulate that a repeat scan in 7–10 days is required when pregnancy is of uncertain viability. However, it is not clear what the repeat scan should show in order to diagnose a miscarriage, and it is commonly accepted that a lack of MSD or CRL growth is a feature of miscarriage. In the second paper by Abdallah et al., data are presented to demonstrate that slow or absent growth of the MSD or CRL is not necessarily associated with miscarriage7. Failure to visualize a yolk sac or embryo on a repeat scan was the defining feature of miscarriage. In the last article in the series, Pexsters and colleagues show that there is significant variation in the accuracy of measurement of CRL and MSD8. In this small, but well-conducted prospective study, variations of up to 20% in measurements were observed between two experienced operators. The study findings suggest that for an MSD of 20 mm, operators may obtain measurements ranging from 16 mm to 24 mm—across the thresholds for diagnosis of miscarriage. From a patient's perspective, if there is one decision a woman would expect clinicians to get right with absolute certainty, it would be whether their baby is alive or not. A cursory internet search will produce a plethora of papers relating to recurrent miscarriage, a problem that affects only 1% of our patient population. Yet where are the large prospective datasets that validate the diagnostic criteria to define miscarriage? There is an urgent need for appropriately powered, prospective studies using agreed reference standards in order to establish the criteria to define miscarriage. Although clinical experience will always be invaluable in caring for individual patients, the evolution of clinical practice into one in which decisions are based on evidence is entirely our responsibility. To think otherwise would be to compromise our professional standing and to do a disservice to our patients. This subject matters, as a single incorrect diagnosis of miscarriage is one too many.