Response to Comment on Witnessing Abusive Head Trauma: Accidents Show Higher Rates of Intracranial Pathologies Than Shaking

I thank Drs. Greeley and Anderst for taking the time to comment1 on my article “Witnessing abusive head trauma: Accidents show higher rates of intracranial pathologies than shaking.”2

The authors raise issues regarding the validity of comparing different data sets. They claim that “pediBIRN is a prospective, cohort study.” In fact, pediBIRN starts with an outcome (child with brain injury), then looks backward (using a clinician's judgment) to decide exposure (accident versus abuse). The PediBIRN authors themselves refer to the data as “strictly observational.”3 So whilst the registry of head injury cases is prosepective, it is not a prospective cohort study of exposures. The studies used in my article all have directionality of outcome to exposure.

Further, the base rate for shaking in my study is “witnessed shaking events,” regardless of which of the four studies4 the data were drawn from. I compiled a list of witnessed cases of shaking from the literature, and the number of such witnessed cases became the base rate in my study. I have not incorporated different base rates from the different studies into my study.

Nevertheless, the authors raise a central, crucial point regarding selection bias. Abusive head trauma (AHT) encompasses a spectrum—from a slap to violent shaking to forceful impact—just as accidental injury ranges from minor bumps to severe falls. Comparing rates of findings in all AHT cases versus all accidents is intractable. Any comparison of rates will be subject to selection bias, particularly selections that influence how severe was the accident and how severe was the abuse.

This is demonstrated in my previous article, “Retino-dural hemorrhages in infants are markers of degree of intracranial pathology not of violent shaking,”5 which shows that when studies claim certain findings are more common in AHT (i.e., specific for AHT), this apparent specificity stems from selection bias. Such studies typically compare severe cases of AHT with less severe accidental injuries. By contrast, when cases are matched for severity (as reflected by hypoxic brain injury), the findings supposedly associated with AHT instead correlate with the degree of brain injury—independent of cause or intent.

The other problem with articles that claim certain findings to be specific to AHT is the methods used for classification of cases as AHT, which are biased towards selecting cases with certain pathologies, creating a self-fulfilling prophecy. The motivation for my article was to restrict comparisons to cases that were witnessed, providing a more robust classification that does not suffer from such circular reasoning.

Unfortunately there is a paucity of such data, and my study required assembling data from different sources. This was explained as a limitation of the study, so I agree with the authors that a degree of caution is warranted in assessing the quoted rates, and I once again reiterate that it will be very helpful to repeat the study with a single high-quality data set.

How then should we interpret these findings from witnessed cases of accidents and shaking? The data clearly show that shaking incidents rarely cause serious injury and typically do not produce the intracranial or ocular findings classically associated with inertial forces or AHT. Indeed, not once has an independently witnessed or videotaped violent shaking of a healthy infant ever resulted in the findings associated with AHT. By contrast, the data shows that independently witnessed accidents (and even videotaped short falls) can cause such findings, even if it is clear that most accidents will not.

It is becoming increasingly clear that results showing that certain findings are specific to AHT, to shaking, or to inertial forces are only found in studies that use circular reasoning6 and/or incorporation bias, and that such results are not reproduced in studies that do not use such deeply flawed methodologies.

Chris B. Brook: conceptualization, writing – original draft.