Intended for healthcare professionals


The evidence base for shaken baby syndrome

BMJ 2004; 328 doi: (Published 25 March 2004) Cite this as: BMJ 2004;328:719
  1. J F Geddes, retired (formerly reader in clinical neuropathology, Queen Mary, University of London) (j.f.geddes{at},
  2. J Plunkett
  1. London
  2. Regina Medical Center, 1175 Nininger Road, Hastings, MN 55033, USA

    We need to question the diagnostic criteria

    The phrase “shaken baby syndrome” evokes a powerful image of abuse, in which a carer shakes a child sufficiently hard to produce whiplash forces that result in subdural and retinal bleeding. The theory of shaken baby syndrome rests on core assumptions: shaking is always intentional and violent; the injury an infant receives from shaking is invariably severe; and subdural and retinal bleeding is the result of criminal abuse, unless proved otherwise.1 These beliefs are reinforced by an interpretation of the literature by medical experts, which may on occasion be instrumental in a carer being convicted or children being removed from their parents. But what is the evidence for the theory of shaken baby syndrome?

    Retinal haemorrhage is one of the criteria used, and many doctors consider retinal haemorrhage with specific characteristics pathognomonic of shaking. However, in this issue Patrick Lantz et al examine that premise (p 754) and conclude that it “cannot be supported by objective scientific evidence.”2 Their study comes hard on the heels of a recently published review of the literature on shaken baby syndrome from 1966 to 1998, in which Mark Donohoe found the scientific evidence to support a diagnosis of shaken baby syndrome to be much less reliable than generally thought.3

    Shaken baby syndrome is usually diagnosed on the basis of subdural and retinal haemorrhages in an infant or young child,1 although the diagnostic criteria are not uniform, and it is not unusual for the diagnosis to be based on subdural or retinal haemorrhages alone.w1 The website of the American Academy of Ophthalmology states that if the retinal haemorrhages have specific characteristics “shaking injury can be diagnosed with confidence regardless of other circumstances.”4 Having reviewed the evidence base for the belief that perimacular folds with retinal haemorrhages are diagnostic of shaking, Lantz et al were able to find only two flawed case-control studies, much of the published work displaying “an absence of… precise and reproducible case definition, and interpretations or conclusions that overstep the data.”2 Their conclusions are remarkably similar to those of Donohoe, who found that “the evidence for shaken baby syndrome appears analogous to an inverted pyramid, with a very small database (most of it poor quality original research, retrospective in nature, and without appropriate control groups) spreading to a broad body of somewhat divergent opinions.”3 His work entailed searching the literature, using the term “shaken baby syndrome” and then assessing the methods of the articles retrieved, using the tools of evidence based inquiry. Reviewing the studies achieving the highest quality of evidence rating scores, Donohoe found that “there was inadequate scientific evidence to come to a firm conclusion on most aspects of causation, diagnosis, treatment, or any other matters,” and identified “serious data gaps, flaws of logic, inconsistency of case definition.”3

    The conclusions of Lantz et al and of Donohoe make disturbing reading, because they reveal major shortcomings in the literature relating to a field in which the opportunity for scientific experimentation and controlled trials does not exist, but in which much may rest on interpretation of the medical evidence.5

    If the concept of shaken baby syndrome is scientifically uncertain, we have a duty to re-examine the validity of other beliefs in the field of infant injury. The recent literature contains a number of publications that disprove traditional expert opinion in the field. A study of independently witnessed low level falls showed that such falls may prove fatal, causing both subdural and retinal bleeding.6 w2 A biomechanical analysis validates that serious injury or death from a low level fall is possible and casts doubt on the idea that shaking can directly cause retinal or subdural haemorrhages.7 w3 An important lucid interval may be present in an ultimately fatal head injury in an infant.8 Neuropathological studies have shown that abused infants do not generally have severe traumatic brain injury and that the structural damage associated with death may be morphologically mild.9 10 What is the relevance of the craniocervical injuries to corticospinal tracts, dorsal nerve roots, and so on that have been described?10 11 We do not know. What is the force necessary to injure an infant's brain? Again, we do not know.

    While most abused children indisputably show the signs of violence, not all do. No one would be surprised to learn that a fall from a two storey building or involvement in a high speed road traffic crash can cause retinal and subdural bleeding, but what is the minimum force required? “It is one thing clearly to state that a certain quantum of force is necessary to produce a subdural hematoma; it is quite another to use examples of obviously extreme force… and then suggest that they constitute the minimum force necessary.”12

    Research in the area of injury to infants is difficult. Quality evidence may need to be based on finite element modelling from data on infants' skulls, brains, and neck structures, rather than living animals. Any studies on immature animal models, if performed, will need to be validated against the known mechanical properties of the human infant. Pending completion of such studies, the reviews by Lantz and Donohoe are a valuable contribution and provide a salutary check for anyone wishing to cite the literature in support of an opinion. Their criticisms of lack of case definition or proper controls can be levelled at the whole literature on child abuse. If the issues are much less certain than we have been taught to believe, then to admit uncertainty sometimes would be appropriate for experts. Doing so may make prosecution more difficult, but a natural desire to protect children should not lead anyone to proffer opinions unsupported by good quality science. We need to reconsider the diagnostic criteria, if not the existence, of shaken baby syndrome.

    Editorial p 720 Clinical review p 754 Letters p 766 Personal view p 775

    Embedded ImageAdditional references w1-w3 are on


    • Competing interests JFG and JP have given evidence in criminal cases at the request of both the prosecution and the defence.


    View Abstract

    Log in

    Log in through your institution


    * For online subscription