Screening for developmental hip disorders has been with us for almost 40 years. Twenty years ago Cliff Roberton famously called it a “mess”.¹ The paper by Roovers et al² published in this issue reports the findings of an observational study to determine whether a strategy of universal ultrasound screening carried out after the newborn period is “more effective” than a strategy based on clinical examination alone. Before considering the implications of this report for policy and practice, it is worth rehearsing some of the challenges facing those who have sought to evaluate this “mess” over the last 40 years.

Developmental dysplasia of the hip (DDH) refers to a spectrum of developmental hip disorders including partial or complete displacement of the femoral head from the acetabulum—that is, developmental displacement of the hip, previously called congenital dislocation of the hip (CDH). Ultrasound imaging of the largely cartilaginous newborn hip, first introduced in the 1980s and pioneered by Graf, has contributed to a paradigm shift in what we mean by CDH. Neonatal hip instability has become synonymous with the outcome it was meant to signal, and dysplastic or shallow hips in the newborn have been added to the disorders for which screening is considered desirable, despite there being only limited understanding of their significance for later hip development.³

When introduced in the 1960s, there was an expectation that early recognition through clinical screening combined with abduction splinting would prevent impaired hip growth and development, serious abnormalities of gait, and premature degenerative changes in the hip joint. To try to prevent these problems by seeking an early diagnosis seems a laudable endeavour. Not all good intentions, however, have their intended consequences.

Sackett et al⁴ suggest questions for clinicians (and policy makers) to ask when deciding whether or not to seek an early diagnosis. To the question about whether the burden of disability from the disease warrants action, the answer for DDH must be “yes”, but only if the action actually helps. Answers to other important questions, including whether early diagnosis through screening really does lead to improved clinical outcomes and whether—at a population level—screening is associated overall with more benefit than harm, are less clear.

The extent to which the expectations of screening programmes have been met has been controversial and difficult to judge, due in part to the absence of a “gold standard” diagnostic test, failure to evaluate the effectiveness of abduction splinting before its widespread introduction, and a paucity of trials or high quality observational studies that have systematically followed screen positive and screen negative children for sufficient periods of time. With these caveats, what has been learnt from observational studies?

In Northern European populations, CDH affects about one in every 1000 children born. We have suggested that in the United Kingdom the number of children requiring at least one operative procedure for CDH has not fallen since screening was introduced,⁵ although this has not been the reported experience of other countries.^6,7 Ultrasound screening programmes for DDH appear to detect a higher proportion of affected children, although false negative diagnoses are not entirely eliminated.⁸ However, this comes at a cost, namely abduction splinting rates of 50–70 per 1000 infants born, compared with an equivalent figure of 4–20 per 1000 for programmes based on clinical screening examination. A small price to pay, it has been argued, given the burden of DDH, except that avascular necrosis of the femoral head, a serious and iatrogenic complication of abduction splinting, affects just under 1% of those treated including those who are unaffected.⁹ Reduction of overtreatment is consequently an important goal to aspire to within universal ultrasound screening programmes.⁸ Hence the study by Roovers et al,² which aims, by screening outside the newborn period, to deliver the benefits of ultrasound in terms of a higher detection rate while avoiding the disbenefits of overtreatment.

In their historical cohort study, a strategy of three ultrasound examinations undertaken at one monthly intervals from 1 month of age is compared with a study of clinical examination carried out in 1993.² The authors report that ultrasound is more “effective” than clinical examination, citing fewer and earlier referrals, a higher detection rate, lower false negative rate, and fewer admissions for inpatient treatment than the clinically screened population. However, despite this, Roovers et al reported that the ultrasound strategy was associated with 4.6% of children receiving treatment, a comparable proportion to that reported from other ultrasound programmes. It would appear then that overtreatment remains an issue even with deferred age at screening.

Drawing inferences from the data presented from this study is complicated by the historical comparison group. As the authors point out, thresholds for hospital admission as well as treatment are likely to have changed quite considerably over the interval between the earlier study of clinical screening and the later study of ultrasound. An additional issue is what is being detected. With ultrasound, 52 per 1000 (most of whom are treated) are reported as affected compared with 35 per 1000 in the earlier clinical study. The diagnostic conundrum rears its head and—somewhat surprisingly and despite an exemplary strategy to follow all screened children to 8 months of age—we are left to infer the false positive rates for each strategy with referral rates—reported as lower in the ultrasound arm—presented as some proxy for these. But is this really the case and is it fair to disregard the costs in their broadest sense of repeated examinations of the initially borderline cases in the community? Clegg et al¹⁰ have succeeded in achieving one of the lowest abduction splinting rates in a universal ultrasound programme by exactly this strategy of repeated ultrasounds for borderline cases. However, this may entail up to eight visits before a treatment decision is made, with some 20% dropping out along the way.

The most scientifically rigorous way to assess the evidence of the effectiveness of a screening policy is a randomised controlled trial (RCT). Trials can and have been conducted in this field. For instance, we and colleagues have evaluated the clinical effectiveness,¹¹ cost effectiveness (A Gray, D Elbourne, C Dezateux et al, unpublished work), and implications for mothers¹² of a policy of ultrasound screening for babies already screened positive on the routine clinical screen. The results suggest that ultrasound imaging used in this way allows the initial screening diagnosis to be refined, leading to a reduction in abduction splinting rates which is not associated with an increase in abnormal hip development or higher rates of operative treatment by 2 years of age, or significantly higher costs or psychosocial problems for mothers. This trial did not, however, consider the role of ultrasound screening for all newborn babies.

Two randomised trials have addressed primary screening.^7,13 Rosendahl et al reported that universal ultrasound was associated with a higher proportion receiving abduction splinting and follow up because of borderline findings, but a lower proportion with “late” subluxation or dislocation. In a subsequent trial, Holen et al⁷ compared universal ultrasound with high quality clinical screening and found that there was no statistically significant difference in the proportion with late presenting hip dysplasia. They recommended using ultrasound only in those with clinical hip instability or recognised risk factors.

If insufficient information is available from RCTs, decision models provide a complementary approach. On the basis of a recent such analysis based largely on observational studies, Dezateux et al⁹ concluded that “ultrasound based screening strategies appeared to be most sensitive and effective but are associated with the greatest risk of potential adverse iatrogenic effects arising in unaffected children.”

So does the paper by Roovers et al² require a rethink of existing screening policy in the United Kingdom?¹⁴ The answer is no, but the challenge facing the United Kingdom will be how to implement high quality clinical screening services and to measure what we do.