The evidence base in child abuse
The editorial entitled “The evidence base in child protection
litigation” BMJ July 22 2006: places a welcome emphasis on the importance
of scientific evidence and specialisation that is required in the clinical
field of child protection. We have undertaken a series of systematic
reviews of the evidence base that underpins the diagnosis of physical
abuse and were surprised, however, that the author describes the evidence
behind child abuse as robust. This broad statement is made on the basis of
16,000 citations for child abuse and historical reference to the formative
work of Tardieu 1860 and Kempe 1962. We feel that this perception is
misleading and fails to assess the quality of the published evidence base.
When a standard systematic approach to the critical appraisal of this
literature is adopted, the findings are very different and there are
significant scientific limitations to many of the published studies.
Child abuse is a generic term and includes physical, sexual, neglect
and emotional abuse. Within each category there are a huge variety of
types of abuse and clinical presentations that result. The
multidisciplinary nature of child protection means that each agency
involved comes with a different agenda and different questions that it
requires the literature to resolve.
As one of the few systematic review teams in this field we have
interrogated the scientific literature around several key questions around
the diagnosis of physical child abuse (www.core-info.cf.ac.uk) . In
general, there was a paucity of literature. Most studies are performed in
the United States and while these studies are invaluable, the difference
in health systems, demographics, definitions and types of abuse and legal
systems mean that the study findings are not necessarily directly
transferable to UK. Studies are compromised by the differences in
definition of abuse used in different countries over the 50 year time
period of this research and the variation in diagnostic techniques. The
security of diagnosis of abuse and reverse causality introduce significant
bias. Case numbers included in studies are small and highly selective and
most observational studies are of a case series design, inherently
compromised by selection bias and lacking any comparative data.
There are areas where the quality of evidence is good, examples
include rib fractures in young children that have a high specificity for
abuse and the profile of non accidental scalds where the literature
defines clear differences between intentional and non intentional scalds.
Our work demonstrates other areas that refute widely established dogma. A
review of 50 years of international literature identified three studies
that give good evidence that the age of bruises cannot be accurately
judged on a visual interpretation of bruise colour . When we evaluated
the evidence surrounding the dating of fractures we identified only three
studies that addressed the topic . These studies included 56 children
under the age of five. They evaluated fracture dating at different
anatomical sites and used different radiological signs to inform their
decision. Each came up with different time frames for these signs of
healing. Radiologists use guidelines drawn together from an individual
expert and published in a textbook: these have yet to be scientifically
validated on a large scale. The overriding feeling is that fractures can
only be aged in the broadest sense.
A torn labial frenum has long been held as a strong indicator of
physical abuse. The entire published evidence around this topic amounts to
28 case reports of torn frenum in abuse. Most cases are of severely abused
children under the age of five. Although accidental torn frena are
mentioned in the literature, there are no published comparative studies.
The probability that a torn frenum is abusive is impossible to calculate
from the literature and a prospective comparative study is called for.
This same profound lack of published evidence applies to the recognition
of adult bites and cigarette burns.
These are a few examples of the strengths and weaknesses in the
scientific literature which lead us to question the term “robust” in
relation to the evidence base.
Towards the end of the article the author reassuringly admits that
there are shortcomings and that the evidence base is a long way from
perfect or complete. He likens the situation to the evidence base for AIDS
or Breast Cancer. There are however profound scientific differences
between these three areas. Both of the latter topics are the subjects of
systematic reviews with meta-analyses that evaluate the strength of
evidence displayed in Randomised Control Trials. Research into the
diagnosis of child abuse cannot avail itself of these trials for obvious
reasons. Studies must rely upon good quality comparative observational
studies. This field of research must be prioritised if we are to promote
the interests of the abused child in the UK in the current climate of
mistrust where clinicians are increasingly reluctant to participate in the
This field of research is challenged in several areas. It is
difficult to perform standard diagnostic studies in the absence of a gold
standard test for abuse that is independent of the presenting injuries or
symptoms of neglect. Consent issues for the inclusion of abused children
and the relative rarity of abuse make it difficult to undertake cohort
studies not to mention the challenges of defining ideal control cases.
David Chadwick is correct that we must be creative in our study design and
use several sources and study types to inform the evidence base that we
draw upon in decision-making.
The overriding priority must be to encourage optimal research in this
field. There are many paediatricians and allied professionals who have
worked in the child protection field for many years and continue to do so.
Many have experience that could be translated into scientific published
evidence if they were able to publish their data from often meticulously
kept retrospective case series or better still use their experience to set
hypothesise and under take well designed prospective comparative studies.
This work urgently needs the support of research funding bodies.
Expert witnesses appearing in court require a thorough understanding
of the quality of the available scientific evidence and must be able to
convey this to the Court in an understandable manner. Courts and
clinicians need to appreciate that child abuse evidence base is a long way
from being robust or complete and that “absence of evidence is not
necessarily evidence of no effect or no association”. Where opinion is
drawn from personal practice, this must be explicit. As Baroness Kennedy
stated ‘A doctor can be convinced, based on his or her experience, that a
defendant is guilty - but unless there is compelling evidence supported
scientifically, he or she should not express that view in criminal
proceedings’ , which sets the standard for an expert opinion in the
Although the clinical field of child protection is going through a
difficult period in the UK, the current situation has the capacity to
stimulate good quality scientific research. This is greatly needed to
build the discipline up to the standards of evidence based clinical
practice that is required throughout clinical practice.
1 Maguire S, Mann M, Sibert J, Kemp A. Can you age bruises accurately
in children? A systematic review. Arch Dis Ch. 2005.90 182-6
2 I.Prosser, S. Maguire, S.K. Harrison, M. Mann, J.R. Sibert, A.M.
Kemp.How old is this fracture? Radiological dating of fractures in
children: A systematic review. AJR; 2005: 184(4)
3 Kennedy H. Sudden Unexpected Death in Infancy. 2004 The Royal
College of Pathologists and The Royal College of Paediatrics and Child
Competing interests: No competing interests