Radiography versus CT for the detection of fractures indicative for child abuse

Radiography versus CT for the

detection of fractures indicative

for child abuse

A review on the accuracy, sensitivity and specificity of

CR and CT for the detection of rib fractures, skull

fractures & classic metaphyseal lesions in young children

Lisa R. Klok, BSc

12403113

Institute of Interdisciplinary Studies

Master Forensic Science

Supervisor:

prof. dr. R.R. van Rijn

Examiner:

prof. dr. R.J. Oostra

Wordcount: 8081, excl. titlepage, bibliography & appendix

Abstract

Fractures are one of the most common clinical findings of non-accidental injury in children. Medical imaging of these fractures and subsequent interpretation can play a pivotal role in the detection of physical abuse. Current imaging protocols prescribe the use of the skeletal survey, a series of radiographs of the whole body. As other imaging techniques are further developed and used in practice, including computed tomography (CT), the use of conventional radiography (CR) for the detection of fractures is debatable. This literature thesis aims to assess the sensitivity and specificity of CT and CR for the detection of fractures indicative of child abuse, namely rib fractures, skull fractures and classical metaphyseal lesions (CMLs). There are sufficient studies that report that CT has a higher sensitivity for the detection of rib and skull fractures than CR, although its specificity is considered lower. Research on the detection of CMLs is limited. An important clinical aspect of CT is the associated radiation dose, but scientists are currently looking into low-dose protocols. For postmortem investigations, the use of CT is advisable as 3D reconstructions and improved fracture dating can be beneficial for further legal proceedings

Keywords:

Abbrevations

CML Classic Metaphyseal Lesion CR Conventional Radiography CT Computed Tomography NAI Non-accidental Injury

Contents iv 1 Introduction 1 1.1 Aim . . . 2 1.2 Study Design . . . 2 2 Rib Fractures 3 2.1 Background Information . . . 3

2.2 Review of Published Literature . . . 4

2.3 Discussion . . . 7

3 Skull Fractures 9

3.1 Background Information . . . 9

3.2 Review of Published Literature . . . 10

3.3 Discussion . . . 11

4 Classic Metaphyseal Lesions 12

4.1 Background Information . . . 12

4.2 Review of Published Literature . . . 13

4.3 Discussion . . . 13 5 Clinical Implications 15 5.1 Radiation Dose . . . 15 6 Legal Implications 17 6.1 Use of 3D Reconstructions. . . 17 6.2 Dating of fractures . . . 17 7 Discussion 18 8 Conclusion 20 Bibliography 21 Appendix 27 A Appendix 27

A.1 PICO Search Strategy . . . 27

Chapter 1

Introduction

It is estimated that in 2017 in the Netherlands, 26 to 37 per 1000 children were maltreated [1]. Especially young children from 0 to 3 years old are at risk of maltreatment, with physical abuse being most prevalent in this age group. An estimated 5.2% of young children are physically abused [1]. The physical trauma due to physical abuse is known as non-accidental injury (NAI). Since fractures are the second most commonly found consequence of NAI [2,3], medical imaging and its interpretation by radiologists can play a pivotal role in the detection of physical abuse [4,5].

Dutch radiologists are required by law to make a formal (skeletal) report when they detect fractures in children under three years old, and when children have died under non-natural circum-stances [6,7]. According to the guidelines of the Dutch expertise centre on child abuse, a skeletal survey must be performed in compliance with standards set by the Royal College of Radiologists & the Society and College of Radiographers [8]. The skeletal survey is a radiological investigation during which the whole body is visualised by detailed radiographic imaging from different angles. For an overview of the required images, see table1.1. The aim of the skeletal survey is not only to detect or exclude fractures but also to date the fractures, to determine possible mechanisms of injury, to differentiate the fractures from normal variants and to diagnose conditions that mimic NAI [9, 10]. Correct diagnosis of NAI is crucial since misdiagnosis can lead to future impairment and even morbidity, while over-diagnosis can have negative impact on family dynamics [9,11,12]. There is continuous research to improve imaging methods to detect NAI accurately. Although conventional radiography (CR) is currently most commonly used to detect fractures, other imaging modalities are still investigated to determine their sensitivity and specificity [9,13]. One of these modalities is computed tomography (CT). CT is already used when intracranial or intra-abdominal injury is suspected due to NAI [14] and increasingly used in postmortem investigation of sudden unexplained death in infants [15]. Developments in CT allow for imaging with lower radiation doses [16–20], reviving the question of whether CT should be standard protocol to image fractures in children. However, there is limited information on the performance of CT for the detection of fractures most predominantly found in child abuse, namely rib fractures, skull fractures and classic metaphyseal lesions [21].

Table 1.1: Skeletal survey guidelines according to the Dutch guidelines. Adapted from [8]

Location Image Additional detailed image

Skull AP

Lateral

Spine Lateral

Abdomen & Pelvis AP (including pelvis)

Upper extremities Humerus AP Elbow lateral Radius & Ulna AP Wrist lateral Lower extremities Femur AP

Knee lateral Knee AP Tibia & Fibula AP Ankle lateral AP: anteroposterior

1.1

Aim

This literature study aims to assess how CT performs in comparison with CR for the detection of skull fractures, rib fractures & metaphyseal corner lesions in abused children. I will also place the performance of CT in its bigger picture, considering relevant doses and legal applications. I believe this will provide a better-substantiated conclusion on the future of CT in suspected child abuse cases in a forensic setting.

1.2

Study Design

For this literature study I reviewed literature published on the accuracy, sensitivity & specificity of the detection of rib fractures, skull fractures and metaphyseal corner lesions in abused children by CR, CT, or both. To find appropriate literature, I searched the PubMed/MEDLINE database for studies published in English and Dutch between January 2000 and December 2019. Additional studies were also included based on recommendations from my supervisor and relevant references in articles previously identified. For the in-depth comparison of CT and CR for the detection of rib fractures, skull fractures and metaphyseal corner lesions I only included observational studies. Furthermore, I excluded reviews, letters, surveys and case reports with five or fewer children. Only studies with children younger than 36 months are included since this is the age group in which the previously mentioned fractures are most prevalent. A complete overview of the search strategy and the results can be found in Appendix A.

Chapter 2

Rib Fractures

2.1

Background Information

Rib fractures are shown to be the most probable fracture found as a result of abuse [2]. Children have a very flexible thorax compared to adults [10]. This flexibility allows for a high degree of elastic and plastic deformation before rib fractures will arise [22]. Therefore, rib fractures are very uncommon in normal handling of children and are often the result of abuse, accidental trauma or (congenital) disorders [10,22]. In more than 80% of the children with rib fractures, the fractures are the result of abuse [23, 24]. Abused children often have multiple rib fractures [23]. A large meta-analysis by Maguire et al. showed that rib fractures in children younger than two years old have a positive predictive value for abuse of 66% [2].

Fractures of the rib in abused children can be the result of anteroposterior compression of the chest, see figure 2.1 [22, 26]. An adult can cause this type of compression if he or she grasps a child around its chest and applies a substantial squeezing force, either in a sudden compression or by violently shaking of the child [22,26, 27]. During anteroposterior compression, the posterior part of the ribs translocates and forms a lever in which the transverse processes of the spine act as the fulcrum [25, 27]. This leverage over the transverse processes causes excessive strain in the posterior fraction of the ribs and can thus lead to medial posterior fractures [27]. Medial posterior fractures can only originate in the absence of a flat hard surface and are therefore not frequently seen in fractures caused by accidental trauma such as a fall [27]. Medial posterior fractures are often grouped with fractures of the posterior arc and labelled as posterior fractures. Posterior fractures are the most common type of rib fractures in children, and between 39% and 44% of all rib fractures are posterior fractures [28]. Fractures in the anterior part of the ribs,

Figure 2.1: Mechanism of injury with anterolateral compression. Typical rib configuration is given in light grey. Anterolateral compression (indicated by the grey arrow) can result in the following fractures (indicated by black arrows): (a) anterior arc fracture, (b) costochondral fracture, (c) lateral arc fracture, (d) posterior arc fracture, (e & f) medial posterior arc fractures. Ad-apted from [25]

Figure 2.2: Healing posterior rib fracture (indicated by arrow) with callus formation in a three-year-old girl in an AP chest radiograph (A) and 3D CT reconstruction (B). Source: [31]

either costochondral or in the anterior arc, are less frequently seen in abused children. It is said that these fractures require larger forces of anterolateral compression [10]. It is thought that other mechanisms of injury, such as direct blows or lateral compression, can also produce anterior fractures [27]. Lateral arc fractures are least seen in abused children, although these might also be produced by severe anteroposterior compression [27].

Since in about one-third of abused children rib fractures are the only pediatric finding [23], it is essential to detect rib fractures accurately. However, radiologists have trouble to detect acute rib fractures on conventional radiographs [29]. Rib fractures tend to displace minimally and might be superimposed over other structures [17,26,30]. Only with healing, and associated callus formation, rib fractures become better identifiable on radiographs [10, 26]. Oblique X-rays and follow-up skeletal surveys reduce these limitations, but CT could also aid in accurate diagnosis of (acute) rib fractures. With CT, early healing stages and surrounding soft tissue injury may be more evident, resulting in higher detection rates of rib fractures. Figure2.2shows an example of a healing rib fracture visualised by CR and CT.

2.2

Review of Published Literature

Studies on the use of CT to detect rib fractures in adults indicate that CT has a higher detection rate than conventional X-rays. A retrospective study by Chapman et al. showed that chest X-rays of blunt force trauma patients missed almost 75% of rib fractures seen on the chest CT scan [32]. Similar studies in adult patients [33], as well as patients younger than 18 years [34], also showed an increased number of detected rib fractures on CT when compared to chest X-rays. However, the reported increase of additional identified rib fractures on CT scans is smaller than in the study of Chapman et al. Since the images in all these studies were made during life, the fracture sites could not be compared to autopsy results or other means. It is therefore unknown whether rib fractures identified on chest X-ray or CT are actual fractures. In contrast, the study by Cattaneo et al. showed different findings [35]. In postmortem imaging of battered pigs, rib fractures detected on X-rays and CT scans were compared to fractures detected by maceration of the skeleton of the piglets. The sensitivity of radiography for the detection of rib fractures was 47%, while the sensitivity of CT scans was 34%.

CHAPTER 2. RIB FRACTURES

One crucial aspect to keep in mind is that these studies looked at fractures produced by blunt force trauma. While this can also be a mechanism of injury in NAI, the anterolateral compression and its related medial posterior fractures are seen more frequently in abused children. Although most studies indicate a higher detection rate of rib fractures on CT, mainly in adults, this does not directly translate to similar detection rates in abused children. In the skeletal surveys, different radiographs are taken compared to the standard chest X-rays in adults, and CT protocols may differ for children. It is therefore still necessary to look at research conducted within the relevant framework to conclude whether CT outperforms CR for the detection of rib fractures in children. Weber et al. looked at the characterization of rib fractures in the postmortem investigation of sudden unexpected death in infancy [36]. In the study, they looked at the reported outcomes of radiographic skeletal surveys by a pediatric radiologist and the reported outcomes of postmortem examinations from autopsies performed by pediatric or forensic pathologists. In 24 children, rib fractures were found during autopsy and included for further analysis. In 67% of the children, fractures were detected accurately on the radiographs. Results show that the detection rate is much higher for healing rib fractures, with 93% of the fractures detected on the skeletal survey, than for acute rib fractures, with a detection rate of 22%. The authors performed no further statistical analysis. Most of these acute fractures are suspected not to be due to child abuse but to be resuscitation-related. It must be considered that fractures due to resuscitation are very rare, and the number in this study may be higher because the study group consisted of children autopsied due to sudden unexpected death in infancy. The study does not indicate whether the detection rate of radiography depends on the location of the fractures. Furthermore, the detection rate is reported as the number of correctly identified children with fractures, not the fractures itself. It is not known whether radiography detected all fractures in these children. The sensitivity of CR for the detection of rib fractures in general may thus be lower than the reported 67%.

A recent retrospective study by Sanchez et al. looked at the detection of rib fractures in living children [17]. They compared results from the initial radiographic skeletal survey to follow-up skeletal survey. The used reference was the total amount of fractures identified on the initial survey and follow-up survey. The study showed that the detection rate of fracture identification of the initial skeletal survey was 84%. There were five children in which no fractures were detected on the initial skeletal survey. These children received additional low-dose CT scans. Eleven fractures were seen on the CT scans and were later also detected on the follow-up skeletal survey. The follow-up skeletal survey also showed CT missed no fractures. The study shows that low-dose CT is an excellent alternative to image acute fractures and fractures in areas that are frequently superimposed on radiographs. It is hard to conclude the true accuracy, sensitivity and specificity of CR and CT based on this study since initial and follow-up skeletal survey also have a limited performance. It does, however, provide support for the use of low-dose CT for the detection of rib fractures within a clinical setting.

A direct comparison study on the performance of CR and CT in abused children who had both was performed by Wootton-Gorges et al. [37]. CR detected only 60% of the fractures seen on CT. CT was significantly better than CR in the detection of posterior and anterior fractures but worse than CR on the detection of lateral fractures. CT also detected significantly more early subacute, subacute and old fractures than CR. CT detected a similar number of acute and late subacute fractures compared to CR. A considerable drawback of this study is that no oblique X-rays were taken, and follow-up skeletal survey were not considered. Both methods increase the accuracy of CR and skeletal survey protocols currently recommended to use both oblique views and follow-up surveys [38]. Since CT and CR were directly compared, it is unknown whether there were fractures missed by both methods. CR identified no new fractures that were missed on CT. Since the reference standard was CT, it is not known whether there were false-positive fractures identified on CT or CR and whether these false positives could have skewed the results. Furthermore, the children were subject to standard CT protocols with a much higher dose than radiographic skeletal surveys protocols.

Figure 2.3: Graphical presentation of the relationship of the sensitivity (A) and specificity (B) of CT and radiography depicted in scatter plots. The points represent the reporters, either registered specialists (red) or consultants (white). Source: [39]

The study by Hong et al. did compare outcomes of CR and CT to autopsy findings [28]. They showed CR had a sensitivity of 28.9% and CT had a sensitivity of 51.5% for the detection of all fractures identified at autopsy during primary investigation. The difference in the detection rate between the two modalities was not statistically significant. Both modalities mainly missed acute fractures, but CT performed slightly better in the detection of these acute fractures. CT also allowed for better detection of fracture healing. There are also indications that CT is better for the detection of anterior fractures than CR.When an experienced ( >20 y.) radiologist reviewed the scans, the sensitivity of CT for the detection of fractures was significantly better than the sensitivity of CR. This might indicate that experience in the analysis of CT scans can benefit accurate detection of rib fractures. Since the study was of retrospective design, identified fractures could not be checked after detection by the radiologists and are thus automatically labelled as false-positives. This study also had a limited number of children with oblique chest radiographs, which limits the sensitivity of CR. Furthermore, the study used a primary investigation team of two radiologist and one experienced radiologist. This study design allows for differences in the study outcome due to inter-observer differences.

Special attention has to be given to the recent study of Shelmerdine et al. [39]. They looked at the sensitivity and specificity of postmortem CR and CT compared to autopsy. It must be noted that 3 of the 25 children in their study were older than 18 months. However, to my current knowledge, it is the most elaborate and statistically reliable study on the performance of CT and CR to date. With the majority of the children within the relevant age range (22 out of 25), it still deserves a prominent spot in this literature review. They had high-quality postmortem CT scans, and the CR images included oblique views. These images were analysed by a multitude of registered specialists and consultants. This study design allowed for statistical analysis with a power of 80% and a significance level of 95%. The gold standard was autopsy results. They found CT had a significantly higher sensitivity for the detection of rib fractures than CR, with a sensitivity of 44.9% and 13.5% respectively. This increased sensitivity can also be seen on figure

2.3A. The specificity of CT was slightly lower than that of CR, with 97.9% and 97.0% respectively. The specificity of the reporters for CT versus radiography is depicted in figure2.3B. It is concluded that:

“by use of CT instead of radiography, an estimated one additional child would be correctly diagnosed as having a rib fracture or fractures for every 6.0 (95% CI 4.5 – 8.7) children that

CHAPTER 2. RIB FRACTURES

would otherwise be diagnosed by use of chest radiographs” Shelmerdine et al. [39] They showed that experience of the reporter did not influence improved sensitivity of CT, in contrast to the conclusion by Hong et al. [28].

2.3

Discussion

As can be seen in table2.1, most studies provide support for the use of CT instead of CR for the detection of rib fractures in children. The sensitivity of CT is higher than that of CR [28,37,39], possibly even with low-dose protocols [17]. CT consistently detects more fractures in almost all locations [28, 37, 39]. CT also provides better detection rates of subtle acute fractures. This might limit the need for additional follow-up imaging of children, meanwhile sending them back to a potentially dangerous environment. It also improves accurate detection and labelling of healing fractures [28,37], which can aid in the determination of NAI. Another advantage of CT is that it also allows for analysis of soft tissue injury and the detection of fractures in the vertebrae [17].

However, there are also severe drawbacks of the studies currently published. Some studies do not compare CT and CR in the same (blinded) manner [17, 36], thus not allowing for direct comparison of these modalities in a reliable way. Wootton Gorges et al. compared CT and CR directly and not to the same reference standard. This only allows for conclusions about the difference between the two modalities, not on their true accuracy [37]. Several studies did use autopsy results as a gold reference to determine the sensitivity and specificity of the modalities [28,39]. Since both studies were of a retrospective design, fractures seen on CR or CT could not be further investigated upon autopsy. New information has shown CT-guided autopsy identifies many fractures upon autopsy that otherwise would have been missed [15]. The labelled false-positive fractures by CT and CR in these studies might thus be an overestimation of the true false-positive number. Furthermore, the study design of Hong and colleagues allowed for inter-observer bias and was statistically underpowered [28].

Before implementation of CT for visualisation of rib fractures in children, more elaborate studies are needed. Many studies do not include oblique chest X-rays in their study design, and the reported sensitivity of CR might thus be an underestimation. The study of Shelmerdine and colleagues provides a good precedent for the use of CT in the postmortem investigation of rib fractures [39]. However, a prospective study design with additional guided autopsy based on fractures detected on CR and CT would be recommended. It is unclear whether the use of CT also benefits the detection of rib fractures in living children. Since it is crucial to keep radiation in these children as low as reasonably achievable, low-dose CT protocols should be applied in living children. The study by Sanchez et al. shows promising results of the use of low-dose CT protocols for the detection of rib fractures [17]. This low-dose CT results in nearly the same effective Table 2.1: Overview of reported performance of CT and CR. For all studies, the rates are given as percentage of fractures identified unless indicated otherwise

Study Tested modality

Gold standard

N (#

fractures) Accuracy Sensitivity Specificity

[36] CR autopsy 24 - 67%a -[37] CR CT 131 - 60%b 100%b [28] CR autopsy 83 - 28.9% 99.9% CT autopsy 101 - 51.5% 99.7% [39] CR autopsy 136 90% b 13.5% (95% CI 8.1 – 21.5) 97.9% (95% CI 96.8 – 98.7) CT autopsy 136 91%b 44.9% (95% CI 31.7 – 58.9) 97.0% (95% CI 95.3 – 98.0) a: sensitivity given per case, not per fracture

radiation dose as most current chest radiography protocols. However, whether the sensitivity of such low-dose CT is significantly higher than the sensitivity than CR for the detection of rib fractures is currently unknown. Postmortem studies that also incorporate low-dose CT protocols could benefit the determination of sensitivity of such low-dose CT scans.

Chapter 3

Skull Fractures

3.1

Background Information

Skull fractures due to abuse are frequently seen in children suffering from abusive head trauma. Abusive head trauma is a very broad term that describes intracranial and spinal lesions due to abuse in children [40]. It often includes severe trauma to the brain and is, therefore, the most serious presentation of physical abuse, with high rates of associated morbidity and mortality [41,42]. In children younger than two years, 79% of the children with abusive head trauma have (complex) skull fractures [43]. The presence of skull fractures in children with abusive head trauma is also a predictor of poor health outcomes [44].

Skull fractures are the result of a large impact force on the head, either by the head hitting something or something hitting the head [22,45]. Hence, skull fractures are also frequently seen as a result of accidental trauma. Nevertheless, since the forces involved in accidental domestic injury, such as falls, are much smaller than in an abusive setting, the manifestation of the skull fractures can be very different. Linear, nondepressed skull fractures are equally common in children suffering from abusive head trauma and accidental injury [40, 46]. In the case of linear fractures, the explanation for the mechanism of injury thus plays a vital role in differentiation between NAI and accidental injury [22]. Complex skull fractures are much more common in NAI, and these include diastatic fractures (widening of the fracture of >1 mm), depressed fractures, a multitude of fractures or fractures crossing suture lines [46]. Although such fractures may be uncommon in accidental injury, accidental injury cannot be excluded. Information on the mechanism of injury and further examination of the child is still necessary to determine NAI [47].

Since abusive head trauma often involves neurological injury, CT is almost always immediately performed if acute head trauma is suspected. CT can quickly image potential soft tissue damage and can thus facilitate urgent medical intervention if needed. Various guidelines recommend the use of CT in suspected abusive head trauma [48, 49]. Most of the CT protocols used to image abusive head trauma also allow for visualization of the bones and subsequent 3D reconstruction of the skull [22,44,50]. However, additional skull radiography is still required as part of the skeletal survey, as NAI can be suspected [44, 51]. Since the benefits of additional skull radiographs are debatable, studies on the sensitivity of both CT and CR for skull fracture detection are very relevant. An example of the visualisation of skull fractures by different techniques can be seen in figure3.1.

Figure 3.1: Lateral skull radiograph (A) and 3D reconstructed CT (B) or a nine-month-old boy. (A) No fractures are seen on the radiograph. (B) A linear skull fracture (arrow) on the right parietal bone with minimal displacement and diastasis was identified after 3D reconstruction of the CT scan. Source: [52]

3.2

Review of Published Literature

Studies in adult trauma victims report a higher sensitivity for the detection of skull fractures for CT than for CR when compared to autopsy results. The sensitivity of (2D)CT for skull fracture detection was between 85 and 89 % and CR had a reported sensitivity of 71 % [53, 54]. Linear fractures can run parallel to the plane of the CT scanner, which can make fractures hard to see on the axial 2D CT slices [55]. High-quality CT allows for further image manipulation, including the reconstruction of 3DCT images. The addition of 3DCT has been shown to have a higher sensitivity for fracture detection than both 2DCT [56–58] as well as radiographs [58–60] in children up to 18 years old.

Culotta et al. directly compared 3DCT scan outcomes to skull radiographs outcomes within a large retrospective study that included 177 children younger than 12 months [52]. In 62 of these children, skull fractures were identified on CR, in 67 children on CT. In six children, fractures were identified on CR that were not seen on CT. Eleven children had skull fractures identified on CT not seen on radiography. The difference between radiography and CT was not significant. 3DCT had a 97% sensitivity and 94% specificity for skull fractures when compared to CR. However, when the false-negative and false-positive CT images were interpreted a second time by experienced radiologists, within most cases the radiologist favoured the outcome of CT above CR. The authors conclude that the elimination of skull radiographs would not influence the detection of skull fractures when the patient has 3DCT imaging available. Since this study directly compared both modalities, statistical interference about the difference in sensitivity & specificity of both modalities was not possible.

A similar comparison study was performed by Sharp and colleagues [51]. They also directly compared skull fracture identification based on CT and skull radiographs in 94 children younger than 23 months. 11 children had fractures seen on CR. All these children’s fractures were also identified on CT. Two more children had fractures identified based on CT, totaling to a total of 13 cases identified on CT. None of the patients had fractures seen on radiography that were not identified on CT. With the small sample size and study design, it is not possible to draw significant conclusions about the added benefit of CR when CT is also conducted.

Prabhu et al. showed the added benefit of 3DCT over radiography and 2DCT in children younger than two years [50]. In 35% of the children with 3D reconstructed models initial CT

CHAPTER 3. SKULL FRACTURES

interpretation was changed. Fractures were labelled in 3DCT as normal variants, and in 1 case a suspected linear fracture was identifiable on 3D reconstruction. The small sample size of this study does not enable statistically significant conclusions about the sensitivity and specificity of the 3D models.

3.3

Discussion

The studies that look into the performance of CT and CR for the detection of skull fractures in children are direct comparison studies [51, 52]. This can be easily explained, since children who are admitted to the hospital with suspected abusive head trauma, are likely to receive both CR and CT. It is valuable to look whether the additional radiation of skull radiographs is necessary in these cases, and direct comparison studies can provide insights into this question. On the other hand, this study design makes it hard to objectively determine which of these imaging modalities performs better for the detection of skull fractures. Systematic studies which use a common reference standard, such as autopsy results, are better suited for this. There are multiple studies on older populations, but they either lack a common reference standard [56, 57, 60] or have a non-existent or severely lacking statistical analysis [53, 58, 59]. Furthermore, CT protocols in younger children might differ due to the size of these children, and because their white matter is unmyelinated [50]. These studies can thus not directly be translated to the younger population.

Not much is known from the perspective of children who have not had head CT but require imaging for the skeletal survey. The radiation dose of standard head CT protocols is significantly higher than that of skull X-ray [51,61]. To my knowledge, there is no research on the possibility of low-dose CT applications for the detection of skull fractures. There are low-dose protocols available with slighter higher effective doses than radiographs [62]. More research is needed to see whether such protocols are appropriate for the detection of skull fractures in children. Since dose is not relevant in postmortem investigations, postmortem CT might be useful for the detection of skull fractures. Since radiographs seem to have little added benefit when CT is performed, head CT might be an appropriate replacement of CR in postmortem investigations. As said before, research in the performance of CT compared to autopsy results is severely lacking. Additional research in the sensitivity & specificity of postmortem CT compared to autopsy standards would be useful. Preferably these studies would have a prospective design so that possible false-positives can be further investigated at autopsy.

Overall, there are strong indications that additional skull radiography does not provide addi-tional benefits if head CT is already performed, as can be seen in table3.1. 3DCT seems preferable over 2DCT and CR for accurate diagnosis of skull fractures. CT could also replace head radio-graphy in postmortem investigations, but more research is needed. CT is currently not a suitable replacement of CR in the skeletal survey of living children due to the associated radiation dose.

Table 3.1: Overview of reported performance of CT and CR. For all studies, the rates are given as percentage of cases identified

Study Tested modality

Gold standard

N (#

cases) Accuracy Sensitivity Specificity

[52] CT CR 177 - 96.8% (95%

CI 88.8 – 96.6)

93.9% (95% CI 87.9 – 97.5)

[51] CR CT 131 - 85%a _100%a

Classic Metaphyseal Lesions

4.1

Background Information

Classic metaphyseal lesions (CML) are fractures that are very specific for child abuse in non-ambulant children [63]. CMLs are fractures through the spongiosa of the metaphysis of long bones. They are often the result of an accumulation of smaller microfractures. Their specificity for child abuse is due to their manner of causation. Large shearing or torsional forces on the metaphysis of immature bones produce CMLs [22]. These kinds of forces are thought to be the result of vigorous pulling and twisting of the limbs or violent shaking [26,30]. Due to the mechanism of injury, these fractures are rarely found after normal handling of non-ambulant, healthy children [22]. Several studies have shown that non-ambulant children cannot generate the required forces themselves and these forces are also not generated in minor accidents as falls [26]. CMLs are most common in the distal femur, distal and proximal tibia and proximal humeri, but also occur in other metaphysis of long bones.

CMLs are also known as corner or bucket handle fractures due to their appearance on radio-graphs, see figure 4.1 [64]. Views perpendicular to the bone often show the well-known bucket fracture, while angled views of CML have a bucket handle appearance. However, CLMs are diffi-cult to identify on radiographs [26]. In CMLs, only a part of the metaphysis is separated from the metaphyseal edge, and the displacement tends to be minimal [26, 31]. Many CMLs heal without callus formation [65]. The lack of the formation of marginal connective tissue and subperiosteal new bone makes CML repair nearly invisible on radiographs. The quick healing time of CMLs, within 4 to 6 weeks, also severely limits the time CMLs are identifiable [66].

Since CT uses the same principles as radiography, the benefits of CT over CR for the detection of CMLs is debated [31, 67]. As CMLs often run along the line of axial CT scans, it is thought that the sensitivity of CT for CMLs is limited [31]. However, 3DCT has already been shown to reduce this problem. Other CT protocols could also be used, such as helical CT. Meanwhile, CT does add much additional radiation. It is still valuable to investigate differences in sensitivity in CT and CR for the detection of CMLs, especially for postmortem investigations.

CHAPTER 4. CLASSIC METAPHYSEAL LESIONS

Figure 4.1: Classical meta-physeal lesion (arrow) visu-alized in different manners. A reconstruction of the pat-tern of a CML can be seen in (A) and (B). (A) A per-pendicular view to the bone shows a corner fracture pat-tern. (B) An angled view shows a bucket handle pat-tern. (C) A radiograph of a 5-month-old child shows a CML in the fibula as a corner fracture. (D) In a radio-graph with different view the same CML shows as a bucket handle. Adapted from [66]

4.2

Review of Published Literature

There is very little recently published literature about the detection of CMLs with either radio-graphy and CT. This may be due to the small target group. Only young children can get CMLs, and it is scarce in healthy children outside of abuse. CMLs used to be visualized mainly in screen-film radiography, but in 2008 Kleinman and colleagues [64] showed that digital radiography had similar performance but allowed for lower radiation dose. Unfortunately, they used simulated frac-tures in pig femurs. Almost all hospitals use digital radiography these days. Tsai et al. [68] also showed that CMLs can be visualised precisely with high-resolution CT. However, these visualisa-tions cannot be translated to standard CT protocols due to the use of micro-CT and the resected specimens.

Tsai and colleagues [65] looked at the value of additional radiography of suspected CML. They determined the prevalence of new bone formation of CML in the distal tibia within their target group based on different radiographic views. They showed that subperiosteal new bone formation was identifiable on 34% of the anteroposterior radiographs. However, when initial lateral and follow-up anteroposterior radiographs are also imaged, prevalence significantly rises to 71%. In almost one-third of the CMLs, no subperiosteal bone formation could be seen. This study does not provide any insights into the accuracy of CR. It does, however, provide strong support for additional radiographic imaging of CMLs by lateral views and follow-up radiography. It also provides insight on the difficulties of CML detection and lack of subperiosteal new bone formation does not exclude the presence of a CML.

4.3

Discussion

As can be seen, there is no recent literature about the use of CT for the detection of CMLs. Since CT has changed a lot since it was first introduced research in its current performance is very valuable. From the current literature, no conclusions can be drawn about the differences between CT and CR for the detection of CMLs. There are indications that CT is not suitable for the

detection of CMLs. The 3DCT reconstruction methods discussed in Chapter 3 ”Skull Fractures” makes it possible to identify fractures parallel to the scanning plane. This is crucial in CML detection, as they often lie in the scanning plane. It is shown in Salter-Harris fractures, fractures of the bone and epiphyses in older children, that 3DCT performs better than CR [69–72]. Since Salter-Harris fractures often also run (partially) in the same plane of the CT scans, this provides support that 3DCT could also be used to detect CMLs. Before CT can be reliably used for the detection of CMLs, it should be proven that CT performs just as well or better than CR. However, recent research is also severely lacking information about the accuracy of CR for the detection of CMLs. Future research should, therefore, also incorporate CR if they want to investigate the use of CT for the detection of CMLs.

Chapter 5

Clinical Implications

5.1

Radiation Dose

Both CR and CT use ionising radiation for image production. One of essential clinical difference between CT and CR is the higher radiation associated with CT. A part of the x-ray radiation used in imaging modalities, such as radiography and CT, is absorbed in the body. This absorbed radiation is called the absorbed dose. Ionising radiation has different effects on different tissues, which is why the ionising effect of imaging is often expressed in its effective dose. The effective dose is an indication of the radiation effects on the whole body by that specific imaging procedure and is expressed in millisievert (mSv) [73, 74]. Radiation can have two types of effects: deterministic and stochastic. The deterministic effects are tissue reactions which occur after threshold dose is crossed. The severity of deterministic effects is often proportional to the radiation dose and can include skin burns and cataracts. The effective radiation required for deterministic effects to occur is considerably greater than the radiation exposure during medical imaging. Stochastic effects occur randomly by chance. Radiation can cause base changes in the DNA, which might be carcinogenic. The probability of these base changes increases with an increase in radiation dose, but the severity of the stochastic effects does not increase.

Radiation is a vital aspect in the imaging of children, as they tend to be more sensitive to ionising radiation and its stochastic effects [73, 74]. As said, some tissues are more susceptible for ionising effects, such as the thyroid, breasts, bone marrow and brain. Due to the size of small children, their organs are closer together, and thus more sensitive organs are either directly irradiated or secondly irradiated by radiation scatter [75]. Due to their growth, children also have higher cell duplication rates. This higher cell duplication rate can cause further duplication of cells with damaged DNA and thus increases the chance on malignancies. Because of their longer life expectancy after imaging, secondary effects such as cancer also have more time to evolve and manifest [76].

Since we cannot neglect these effects of CR and CT, it is crucial to consider both the potential drawbacks of imaging and not imaging children who might suffer from NAI. There is consensus within the medical community that the skeletal surveys remain necessary to identify signs of NAI. In one-third of the skeletal surveys of children with suspected abuse, fractures are detected [42]. Meanwhile, a study estimated a relatively low lifetime risk of developing cancer due to medical imaging in children [77]. They calculated an additional risk of 0.016% per millisievert for developing cancer. To put the effective dose of radiation from CR and CT in perspective, some estimations of the radiation per procedure are given in table 5.1. The background radiation is estimated to have an effective dose between 2.7 and 3.5 mSv/year [74,78].

Based on radiation and its effects, careful consideration is required before CT can replace CR. The added benefits of CT should weigh up to the associated increased radiation and therefore, the increase in risk of cancer later in life. From this perspective, CT should only replace CR when the sensitivity of CT is significantly better than that of CR with minimal added radiation.

Table 5.1: Effective radiation dose of different procedures. Estimation for children younger than 2.5 years.

Procedure Estimated effective dose (mSv) Source SS according to RCPCH 0.9 – 1.8 [73]

4-view chest radiograph 0.35 [17]

Low-dose chest CT 0.48 [17]

2-way head radiograph 0.04 [51]

Head CT 1.2 - 2 [51]

Specific protocols can lower the radiation of CT, or the sensitivity of CT might limit the need follow-up imaging [17], thereby also decreasing radiation dose. Until such claims can be sufficiently proven, CR is preferable over CT radiation wise. However, in clinical settings, CR is often still performed after CT. Since the goal of medical professionals is to keep radiation as low as reasonably achievable, this seems unnecessary. Studies have shown good indications CT scans of the skull and chest/abdomen have a similar or greater detection for fractures rate as CR [28,37,39,51,52]. Is seems unreasonable to perform additional CR in these situations, even when the additional radiation exposure is relatively low.

Chapter 6

Legal Implications

6.1

Use of 3D Reconstructions

As we could see in Chapter 3 ”Skull Fractures”, 3DCT improves the sensitivity of CT for the detection of skull fractures. Another benefit of 3DCT is in the use of 3DCT within legal procedures. Not only is 3DCT better to judge by radiologists [58], but it can also help the understanding of medical layman [26,58]. 3DCT provides a better representation, while 2D scans and radiography require more imagination and training. 3DCT can also help to visualize potential mechanisms of injury in models [79,80].

Since suspected cases of NAI often require further consultation with youth services and some-times even further legal action, 3DCT could be advantageous. It remains important that the 3D reconstructions remain explained by expert and are not misinterpreted [80].

6.2

Dating of fractures

Within the medico-legal field, the dating of fractures in children can be of importance to either col-laborate or contest statements made by the caregivers. However, there are considerable variations in the healing process of the bone, and it is difficult to assess the healing stage of the fracture on medical images. These aspects can lead to large uncertainty in the estimation of the timing of the injury. This large uncertainty can be challenging to understand for medical laypeople, who prefer more defined terms [81]. As was shown in Chapter 2 “Rib Fractures”, several studies showed CT allowed for better labelling of the healing stage of fractures. This improved classification of fracture healing allows for a narrower estimation of the age of a fracture and may thus improve evidence interpretation by laypeople such as judges and juries. Furthermore, if multiple stages of healing can be identified within one child, this provides strong support for NAI [81]. The use of CT might thus be helpful in child abuse investigations by easing the classification of fracture healing stages and narrowing fracture age estimations.

Discussion

For this literature study I had to focus on specific aspects of the medical imaging of fractures in children that I wanted to highlight. Firstly, I choose to concentrate on the performance of CT and CR. There are many other techniques possible to detect fractures, like Magnetic Resonance Imaging (MRI), ultrasound and Positron Emission Tomography (PET) scanning. I chose to focus on CT and CR since these modalities are most commonly used in paediatrics and have commonly available protocols that can be applied by most hospitals. Whole-body MRI is also promising for fracture detection without the use of ionising radiation [82]. However, whole-body MRI still has a relatively long scanning time, and thus often requires sedation. With the development of rapid MRI protocols [83], this modality may play an important role in the future. However, there is a lot of additional research required before these modalities are also applicable for fracture detection on a larger scale.

Furthermore, an important aspect of an imaging technique is also its ability to identify possible differential diagnoses. Underlying pathologies may cause fractures also seen in NAI. Fractures due to NAI visualised on medical images should be differentiated from normal variants or underlying pathologies. In-depth information of medical mimics of NAI is outside the scope of this study, but more information can be found in the study of Christian et al. [84].

As can be seen, the literature published in the last twenty years about the performance of CT and CR for fracture detection is quite limited. Studies often refer to older papers (before 2000 A.D.), but CT and radiography have developed a lot over recent years. In the field of radiography, digital radiography has completely replaced the traditional screen-film [66]. Since 2000, multislice CT is commonly used and multiple image reconstruction algorithms, including Filtered Back Projection based algorithms and iterative reconstruction were introduced [85]. Because of these developments, older research may not be a good representation of the applications of current CR and CT.

Moreover, most of the recently published papers have a small sample size or have a descriptive character and thus lack a thorough statistical analysis. Because the studies are so different in study design, it is difficult to merge their data or perform a more extensive meta-analysis. The studies often lack interpretation by a multitude of specialists. It has been shown that accuracy of imagining modalities is dependent on the observer [29]. Studies that use a limited number of observers can thus have an inherent observer bias in their reported outcomes, and their reported accuracy may not be a good representation for the accuracy of the method for all specialists.

Finally, different studies also often use different gold standards. Studies in living children are often direct comparison studies, which is logical given their study design. However, direct comparison studies only allow for conclusions about the differences between CT and CR and which modality has a higher sensitivity or specificity cannot be concluded. A reliable gold standard should be used to determine sensitivity and specificity of medical imaging modalities. Most studies use autopsy as their gold standard. Although autopsy is undoubtedly a useful reference, the autopsy itself may also be flawed. Autopsy can miss fractures that might be detected with CT and CR [15,86]. In studies with a retrospective design, autopsy findings cannot be checked after

CHAPTER 7. DISCUSSION

imaging interpretation. As a result, fractures might have been missed at autopsy but present on CR, on CT or both. This can cause an overestimation of the false positive number and thus a decrease in the reported specificity. Prospective studies would be valuable to determine sensitivity and specificity in the future. Areas of suspected fractures on CR or CT could be carefully checked within the body and sent in for further histological examination to determine the presence of a fracture. Another option for a gold standard is the maceration of the body or suspected bones and subsequent fracture determination [35]. However, since studies involve the bodies of children, this raises many ethical concerns.

In summary, there is a lack of large-scale studies that uses multiple observers that look in fracture detection with CT and CR. Hopefully, the recent study of Shelmerdine et al.[39] has inspired other researchers to perform similar studies for other fracture types or low-dose CT protocols. Preferentially, some of the future studies are of a prospective design and allow to check suspected fractures further. Large-scale studies allow us to get a better understanding of the sensitivity and specificity of the modalities. This knowledge can be very insightful to establish future protocols when NAI is suspected. It remains imperative to be as accurate as possible when fractures are present as a result of NAI.

Conclusion

Before CT should be considered as a replacement of CR in live children, more research is required. The additional radiation of CT may not weigh up against its benefits. More low-dose protocols should be developed, and their potential should be researched. However, the use of additional radiography when a head CT or chest CT is already acquired seems overabundant. CT seems to perform just as well or even better than CR for the detection of rib and skull fractures, and it would limit further radiation exposure of children. There is currently no research available on the performance of CT for the detection of CMLs.

For postmortem investigations, it is advisable to perform CR as well as CT. This structure will allow for further studies into the sensitivity and specificity. There are sufficient indications CT sensitivity is better than that of CR, although its specificity is reportedly lower. Guided autopsy based on fractures identified on CT can give us better insight into its false positive-rate in the future. The use of CT allows for 3D reconstructions of the CT scans and improved dating of fractures. These factors can be beneficial for further legal proceedings in postmortem investigations.

Bibliography

[1] L. Alink, M. Prevoo, S. van Berkel, M. Linting, M. Klein Velderman, and F. Pannebakker. NPM-2017: Nationale Prevalentiestudie Mishandeling van Kinderen en Jeugdigen. Technical report, Universiteit Leiden/ TNO, Leiden, 2018. 1

[2] S. Maguire, L. Cowley, M. Mann, and A. Kemp. What does the recent literature add to the identification and investigation of fractures in child abuse: an overview of review updates 2005-2013. Evidence-Based Child Health: A Cochrane Review Journal, 8(5):2044–2057, sep 2013. 1,3

[3] P. McMahon, W. Grossman, M. Gaffney, and C. Stanitski. Soft-tissue injury as an indication of child abuse. Journal of Bone and Joint Surgery - Series A, 77(8):1179–1183, 1995. 1

[4] N. Jain. The role of diagnostic imaging in the evaluation of child abuse. BCMJ, 57(8):336–340, 2015. 1

[5] R. R. Van Rijn, N. Kieviet, R. Hoekstra, H. G. T. Nijs, and R. A. C. Bilo. Radiology in suspected non-accidental injury: theory and practice in The Netherlands. European journal of radiology, 71(1):147–151, jul 2009. 1

[6] Besluit verplichte meldcode huiselijk geweld en kindermishandeling, 2013. 1

[7] Ministerie van Justitie en Veiligheid. Geweld hoort nergens thuis. Technical report, 2018. 1

[8] Landelijk Expertise Centrum Kindermishandeling (LECK). Protocol Skeletstatus. Technical Report september 2018, 2018. 1,2

[9] B. Karmazyn, E. M. Miller, S. E. Lay, J. M. Massey, M. R. Wanner, M. B. Marine, S. G. Jennings, F. Ouyang, and R. A. Hibbard. Double-read of skeletal surveys in suspected non-accidental trauma: what we learned. Pediatric radiology, 47(5):584–589, may 2017. 1

[10] Robert A C Bilo, Rick R van Rijn, and Simon G Robben. Forensische aspecten van fracturen op de kinderleeftijd. Isala series, 2009. 1,3,4

[11] H. Carty and A. Pierce. Non-accidental injury: A retrospective analysis of a large cohort. European Radiology, 12(12):2919–2925, dec 2002. 1

[12] R. D. Powell-Doherty, N. E. Raynor, D. A. Goodenow, D. G. Jacobs, and A. Stallion. Ex-amining the role of follow-up skeletal surveys in non-accidental trauma. American Journal of Surgery, 213(4):606–610, apr 2017. 1

[13] A. Nguyen and R. Hart. Imaging of non-accidental injury; what is clinical best practice? Journal of medical radiation sciences, 65(2):123–130, jun 2018. 1

[14] G. Partan, P. Pamberger, E. Blab, and W. Hruby. Common tasks and problems in paediatric trauma radiology. European journal of radiology, 48(1):103–124, oct 2003. 1

[15] K. W. Feldman. Rib fractures: elusive, but important. The Lancet. Child & adolescent health, 2(11):769–770, nov 2018. 1,7,18

[16] T. R. Sanchez, J. S. Lee, K. P. Coulter, J. A. Seibert, and R. Stein-Wexler. CT of the chest in suspected child abuse using submillisievert radiation dose. Pediatric radiology, 45(7):1072– 1076, jul 2015. 1

[17] T. R. Sanchez, A. D. Grasparil, R. Chaudhari, K. P. Coulter, and S. L. Wootton-Gorges. Characteristics of Rib Fractures in Child Abuse-The Role of Low-Dose Chest Computed Tomography. Pediatric emergency care, 34(2):81–83, feb 2018. 4,5,7,16

[18] M. Mahesh. Advances in CT technology and application to pediatric imaging., 2011. [19] M. I. Karamat. Strategies and Scientific Basis of Dose Reduction on State-of-the-Art Multirow

Detector X-Ray CT Systems. Critical reviews in biomedical engineering, 43(1):33–59, 2015. [20] C. Lee, M. S. Pearce, J. A. Salotti, R. W. Harbron, M. P. Little, K. McHugh, C. L. Chapple,

and A. B. De Gonzalez. Reduction in radiation doses from paediatric CT scans in Great Britain. British Journal of Radiology, 89(1060), 2016. 1

[21] K. Nimkin and P. K. Kleinman. Imaging of child abuse. Radiologic clinics of North America, 39(4):843–864, jul 2001. 1

[22] M. Paddock, A. Sprigg, and A. C. Offiah. Imaging and reporting considerations for sus-pected physical abuse (non-accidental injury) in infants and young children. Part 1: initial considerations and appendicular skeleton. Clinical radiology, 72(3):179–188, mar 2017. 3, 9,

12

[23] K. A. Barsness, E. S. Cha, D. D. Bensard, C. M. Calkins, D. A. Partrick, F. M. Karrer, and J. D. Strain. The positive predictive value of rib fractures as an indicator of nonaccidental trauma in children. Journal of Trauma, 54(6):1107–1110, 2003. 3,4

[24] B. Bulloch, C. J. Schubert, P. D. Brophy, N. Johnson, M. H. Reed, and R. A. Shapiro. Cause and clinical characteristics of rib fractures in infants. Pediatrics, 105(4), 2000. 3

[25] P. K. Kleinman and P. D. Barnes. Head trauma. In Diagnostic Imaging of Child Abuse, pages 285–342. Mosby, St. Louis, 2 edition, 1998. 3

[26] J. R. Dwek. The radiographic approach to child abuse. Clinical orthopaedics and related research, 469(3):776–789, mar 2011. 3,4,12,17

[27] M. J. Worn and M. D. Jones. Rib fractures in infancy: establishing the mechanisms of cause from the injuries–a literature review. Medicine, science, and the law, 47(3):200–12, jul 2007.

3,4

[28] T. S. Hong, J. A. Reyes, R. Moineddin, D. A. Chiasson, W. E. Berdon, and P. S. Babyn. Value of postmortem thoracic CT over radiography in imaging of pediatric rib fractures. Pediatric radiology, 41(6):736–748, jun 2011. 3,6,7,16

[29] A. C. Offiah, L. Moon, C. M. Hall, and A. Todd-Pokropek. Diagnostic accuracy of fracture detection in suspected non-accidental injury: the effect of edge enhancement and digital display on observer performance. Clinical radiology, 61(2):163–173, feb 2006. 4,18

[30] E. Raynor, P. Konala, and A. Freemont. The detection of significant fractures in suspected infant abuse. Journal of forensic and legal medicine, 60:9–14, nov 2018. 4,12

[31] R. R. Van Rijn and T. Sieswerda-Hoogendoorn. Educational paper: imaging child abuse: the bare bones. European journal of pediatrics, 171(2):215–224, feb 2012. 4,12

[32] B. C. Chapman, D. M. Overbey, F. Tesfalidet, K. Schramm, R. T. Stovall, A. French, J. L. Johnson, C. C. Burlew, C. Barnett, E. E. Moore, and F. M. Pieracci. Clinical Utility of Chest Computed Tomography in Patients with Rib Fractures CT Chest and Rib Fractures. Archives of Trauma Research, 5(4), sep 2016. 4

BIBLIOGRAPHY

[33] M. Traub, M. Stevenson, S. McEvoy, G. Briggs, S. K. Lo, S. Leibman, and T. Joseph. The use of chest computed tomography versus chest X-ray in patients with major blunt trauma. Injury, 38(1):43–47, 2007. 4

[34] J. Renton, S. Kincaid, and P. F. Ehrlich. Should helical CT scanning of the thoracic cavity replace the conventional chest x-ray as a primary assessment tool in pediatric trauma? An efficacy and cost analysis. Journal of pediatric surgery, 38(5):793–797, may 2003. 4

[35] C. Cattaneo, E. Marinelli, A. Di Giancamillo, M. Di Giancamillo, O. Travetti, L. Vigano’, P. Poppa, D. Porta, A. Gentilomo, and M. Grandi. Sensitivity of autopsy and radiological examination in detecting bone fractures in an animal model: implications for the assessment of fatal child physical abuse. Forensic science international, 164(2-3):131–137, dec 2006. 4,

19

[36] M. A. Weber, R. A. Risdon, A. C. Offiah, M. Malone, and N. J. Sebire. Rib fractures identified at post-mortem examination in sudden unexpected deaths in infancy (SUDI). Forensic science international, 189(1-3):75–81, aug 2009. 5,7

[37] S. L. Wootton-Gorges, R. Stein-Wexler, J. W. Walton, A. J. Rosas, K. P. Coulter, and K. K. Rogers. Comparison of computed tomography and chest radiography in the detection of rib fractures in abused infants. Child abuse & neglect, 32(6):659–663, jun 2008. 5,7,16

[38] A. Offiah, R. R. van Rijn, J. M. Perez-Rossello, and P. K. Kleinman. Skeletal imaging of child abuse (non-accidental injury). Pediatric radiology, 39(5):461–470, may 2009. 5

[39] S. C. Shelmerdine, D. Langan, J. C. Hutchinson, M. Hickson, K. Pawley, J. Suich, L. Palm, N. J. Sebire, A. Wade, and O. J. Arthurs. Chest radiographs versus CT for the detection of rib fractures in children (DRIFT): a diagnostic accuracy observational study. The Lancet. Child & adolescent health, 2(11):802–811, nov 2018. 6,7,16,19

[40] A. K. Choudhary, S. Servaes, T. L. Slovis, V. J. Palusci, G. L. Hedlund, S. K. Narang, J. A. Moreno, M. S. Dias, C. W. Christian, M. D. Jr Nelson, V. M. Silvera, S. Palasis, M. Raissaki, A. Rossi, and A. C. Offiah. Consensus statement on abusive head trauma in infants and young children. Pediatric radiology, 48(8):1048–1065, aug 2018. 9

[41] C. J. Hobbs. ABC of child abuse. Head injuries. BMJ, 298(6681):1169–1170, apr 1989. 9

[42] A. M. Kemp, F. Dunstan, S. Harrison, S. Morris, M. Mann, K. Rolfe, S. Datta, D. P. Thomas, J. R. Sibert, and S. Maguire. Patterns of skeletal fractures in child abuse: Systematic review, oct 2008. 9,15

[43] V. J. Palusci. Risk factors and services for child maltreatment among infants and young children. Children and Youth Services Review, 33(8):1374–1382, aug 2011. 9

[44] S. Rajaram, R. Batty, C. D. C. Rittey, P. D. Griffiths, and D. J. A. Connolly. Neuroimaging in non-accidental head injury in children: an important element of assessment. Postgraduate medical journal, 87(1027):355–361, may 2011. 9

[45] J. M. Hardman. The pathology of traumatic brain injuries. Advances in neurology, 22:15–50, 1979. 9

[46] C. J. Hobbs. Skull fracture and the diagnosis of abuse. Archives of Disease in Childhood, 59(3):246–252, 1984. 9

[47] N. Stoodley. Neuroimaging in non-accidental head injury: if, when, why and how. Clinical radiology, 60(1):22–30, jan 2005. 9

[48] Head injury: Triage, assessment, investigation and early management of head injury in chil-dren, young people and adults. Technical report, National Clinical Guideline Centre (UK), London, 2014. 9

[49] The radiological investigation of suspected physical abuse in children. Technical report, The royal college of Radiologists, 2018. 9

[50] S. P. Prabhu, A. W. Newton, J. M. Perez-Rossello, and P. K. Kleinman. Three-dimensional skull models as a problem-solving tool in suspected child abuse. Pediatric radiology, 43(5):575– 581, mar 2013. 9,10,11

[51] S. R. Sharp, S. M. Patel, R. E. Brown, and C. Landes. Head imaging in suspected non accidental injury in the paediatric population. In the advent of volumetric CT imaging, has the skull X-ray become redundant? Clinical radiology, 73(5):449–453, may 2018. 9, 10,11,

16

[52] P. A. Culotta, J. E. Crowe, Q. Tran, J. Y. Jones, A. R. Mehollin-Ray, H. B. Tran, M. Donaruma-Kwoh, C. T. Dodge, E. A. Camp, and A. T. Cruz. Performance of computed tomography of the head to evaluate for skull fractures in infants with suspected non-accidental trauma. Pediatric radiology, 47(1):74–81, jan 2017. 10,11,16

[53] H. Chawla, R. K. Yadav, M. S. Griwan, R. Malhotra, and P. K. Paliwal. Sensitivity and spe-cificity of CT scan in revealing skull fracture in medico-legal head injury victims. Australasian Medical Journal, 8(7):235–238, aug 2015. 10,11

[54] H. Chawla, R. Malhotra, R. K. Yadav, M. S. Griwan, P. K. Paliwal, and A. D. Aggarwal. Dia-gnostic utility of conventional radiography in head injury. Journal of Clinical and DiaDia-gnostic Research, 9(6):TC13–TC15, jun 2015. 10

[55] D. M. Fleece and P. S. Kochan. Skull Fracture in an Infant Not Visible with Computed Tomography. Journal of Pediatrics, 154(6):934, jun 2009. 10

[56] G. Orman, M. W. Wagner, D. Seeburg, C. A. Zamora, A. Oshmyansky, A. Tekes, A. Poretti, G. I. Jallo, T. A. G. M. Huisman, and T. Bosemani. Pediatric skull fracture diagnosis: should 3D CT reconstructions be added as routine imaging? Journal of Neurosurgery: Pediatrics, 16(4):426–431, oct 2015. 10,11

[57] S. Y. Sim, H. G. Kim, S. H. Yoon, J. W. Choi, S. M. Cho, and M. S. Choi. Reappraisal of Pediatric Diastatic Skull Fractures in the 3-Dimensional CT Era: Clinical Characteristics and Comparison of Diagnostic Accuracy of Simple Skull X-Ray, 2-Dimensional CT, and 3-Dimensional CT. World Neurosurgery, 108:399–406, dec 2017. 11

[58] M. T. Parisi, R. T. Wiester, S. L. Done, N. F. Sugar, and K. W. Feldman. Three-Dimensional Computed Tomography Skull Reconstructions as an Aid to Child Abuse Evaluations. Pedi-atric emergency care, 31(11):779–786, nov 2015. 10,11,17

[59] M. H. Mulroy, A. M. Loyd, D. P. Frush, T. G. Verla, B. S. Myers, and C. R. ’Dale’ Bass. Evaluation of pediatric skull fracture imaging techniques. Forensic Science International, 214(1-3):167–172, jan 2012. 11

[60] Y. Kim, J. Cheong, and S. H. Yoon. Clinical comparison of the predictive value of the simple skull x-ray and 3 dimensional computed tomography for skull fractures of children. Journal of Korean Neurosurgical Society, 52(6):528–33, dec 2012. 10,11

[61] W. E. Berdon and K. W. Feldman. A modest proposal: thoracic CT for rib fracture diagnosis in child abuse., feb 2012. 11

[62] S. E. J. Connor, T. Arscott, J. Berry, L. Greene, and R. O’Gorman. Precision and accuracy of low-dose CT protocols in the evaluation of skull landmarks. Dentomaxillofacial Radiology, 36(5):270–276, jul 2007. 11

BIBLIOGRAPHY

[63] P. K. Kleinman, J. M. Perez-Rossello, A. W. Newton, H. A. Feldman, and P. L. Kleinman. Prevalence of the classic metaphyseal lesion in infants at low versus high risk for abuse. AJR. American journal of roentgenology, 197(4):1005–1008, oct 2011. 12

[64] P. L. Kleinman, D. Zurakowski, K. J. Strauss, R. H. Cleveland, J. M. Perez-Rosello, D. P. Nichols, K. H. Zou, and P. K. Kleinman. Detection of Simulated Inflicted Metaphyseal Fractures in a Fetal Pig Model: Image Optimization and Dose Reduction with Computed Radiography 1. Radiology, 247(2), 2008. 12,13

[65] A. Tsai, S. A. Connolly, K. Ecklund, P. R. Johnston, and P. K. Kleinman. Subperiosteal new bone formation with the distal tibial classic metaphyseal lesion: prevalence on radiographic skeletal surveys. Pediatric radiology, 49(4):551–558, apr 2019. 12,13

[66] P. K. Kleinman. Problems in the diagnosis of metaphyseal fractures. Pediatric radiology, 38 Suppl 3:S388–94, jun 2008. 12,13,18

[67] R. R. Van Rijn. How should we image skeletal injuries in child abuse? Pediatric radiology, 39 Suppl 2:S226–9, apr 2009. 12

[68] A. Tsai, A. G. McDonald, A. E. Rosenberg, R. Gupta, and P. K. Kleinman. High-resolution CT with histopathological correlates of the classic metaphyseal lesion of infant abuse. Pedi-atric radiology, 44(2):124–140, feb 2014. 13

[69] S. P. Lemburg, E. Lilienthal, and C. M. Heyer. Growth plate fractures of the distal tibia: Is CT imaging necessary? Archives of Orthopaedic and Trauma Surgery, 130(11):1411–1417, nov 2010. 14

[70] A. Nenopoulos, T. Beslikas, I. Gigis, F. Sayegh, I. Christoforidis, and I. Hatzokos. The role of CT in diagnosis and treatment of distal tibial fractures with intra-articular involvement in children. Injury, 46(11):2177–2180, nov 2015.

[71] D. Thawrani, V. Kuester, P. G. Gabos, R. W. Kruse, A. G. Littleton, K. J. Rogers, L. Holmes, and M. M. Thacker. Reliability and necessity of computerized tomography in distal tibial physeal injuries. Journal of Pediatric Orthopaedics, 31(7):745–750, oct 2011.

[72] W. C. Lippert, R. F. Owens, and E. J. Wall. Salter-Harris type III fractures of the distal femur plain radiographs can be deceptive. Journal of Pediatric Orthopaedics, 30(6):598–605, sep 2010. 14

[73] M. Bajaj and A. C. Offiah. Imaging in suspected child abuse: necessity or radiation hazard? Archives of disease in childhood, 100(12):1163–1168, dec 2015. 15,16

[74] T. L. Slovis, P. J. Strouse, and K. J. Strauss. Radiation Exposure in Imaging of Suspected Child Abuse: Benefits versus Risks. The Journal of pediatrics, 167(5):963–968, nov 2015. 15

[75] SOURCES AND EFFECTS OF IONIZING RADIATION, Vol. 1. Technical report, UN-SCEAR, New York, 2000. 15

[76] K. Ozasa, Y. Shimizu, A. Suyama, F. Kasagi, M. Soda, E. J. Grant, R. Sakata, H. Sugiyama, and K. Kodama. Studies of the Mortality of Atomic Bomb Survivors, Report 14, 1950–2003: An Overview of Cancer and Noncancer Diseases. Radiation Research, 177(3):229–243, mar 2012. 15

[77] J. Van Aalst, C. R. L. P. N. Jeukens, J. S. H. Vles, E. A. Van Maren, A. G. H. Kessels, D. L. H. M. Soudant, J. W. Weber, A. A. Postma, and E. M. J. Cornips. Diagnostic radiation exposure in children with spinal dysraphism: An estimation of the cumulative effective dose in a cohort of 135 children from the Netherlands. Archives of Disease in Childhood, 98(9):680– 685, sep 2013. 15

[78] Ionising radiation: dose comparisons - GOV.UK. 15

[79] D. Wittschieber, L. Beck, V. Vieth, and M. L. Hahnemann. The role of 3DCT for the eval-uation of chop injuries in clinical forensic medicine. Forensic Science International, 266:e59– e63, sep 2016. 17

[80] A. Borowska-Solonynko and B. Solonynko. The use of 3D computed tomography reconstruc-tion in medico-legal testimony regarding injuries in living victims - Risks and benefits. Journal of Forensic and Legal Medicine, 30:9–13, 2015. 17

[81] T. A. Pickett. The challenges of accurately estimating time of long bone injury in children, jul 2015. 17

[82] S. F. Kralik, N. Supakul, I. C. Wu, G. Delso, R. Radhakrishnan, C. Y. Ho, and K. A. Eley. Black bone MRI with 3D reconstruction for the detection of skull fractures in children with suspected abusive head trauma. Neuroradiology, 61(1):81–87, jan 2019. 18

[83] H. Mehta, J. Acharya, A. L. Mohan, M. E. Tobias, L. LeCompte, and D. Jeevan. Minim-izing radiation exposure in evaluation of pediatric head trauma: Use of rapid MR imaging. American Journal of Neuroradiology, 37(1):11–18, jan 2016. 18

[84] C. W. Christian and L. J. States. Medical Mimics of Child Abuse. AJR. American journal of roentgenology, 208(5):982–990, may 2017. 18

[85] X. Pan, J. Siewerdsen, P. J. La Riviere, and W. A. Kalender. Anniversary paper: Development of x-ray computed tomography: The role of Medical Physics and AAPM from the 1970s to present, 2008. 18

[86] E. P. McGraw, J. E. Pless, D. J. Pennington, and S. J. White. Postmortem radiography after unexpected death in neonates, infants, and children: should imaging be routine? AJR. American journal of roentgenology, 178(6):1517–1521, jun 2002. 18

Appendix A

Appendix

A.1

PICO Search Strategy

First to identify relevant search terms, I used the PICO strategy, a commonly used strategy in evidence-based medicine. For an overview of the used PICO search strategy, see figureA.1. Research question: Is Computed Tomography an alternative of Conventional Radiography for the detection of skull fractures, (posterior) rib fractures metaphyseal corner fractures in abused children?

Relevant time frame: CT developed over the last decades. Since 2000, multislice CT is com-monly used and multiple image reconstruction algorithms, including Filtered Back Projection based algorithms, were introduced1_{. Therefore, literature is limited to studies after 2000.}

Not included aspects in research question: Forensic relevance, Prevalence, influence of dose, current practices and psychological aspects, potential other imaging modalities/ methods. Additional terms: Forensic, dose, skeletal status.

1_{Pan, X., Siewerdsen, J., La Riviere, P., Kalender, W. (2008). Anniversary Paper: Development of x-ray} computed tomography: The role of Medical Physics and AAPM from the 1970s to present. Medical Physics, 35(8), 3728-3739. doi: 10.1118/1.2952653

Figure A.1: PICO search strategy including relevant terms.

A.2

PubMed/ MEDLINE Search outcome

Articles interesting for whole review, not only comparison:

Search strategy Number of articles

(”Child Abuse”[Mesh]) AND

(((”Radiography”[Mesh]) OR ”Tomography, X-Ray Computed”[Mesh]) OR ”imaging” AND Fracture*)

547 * Filters activated:

Publication date from 2000/01/01 to 2019/12/31, Dutch, English 314 * Selection based on title

(exclusion of non-relevant subjects, letters) 163

* Selection based on abstract 67

* Selection based on article 58

Inclusion of articles recommended by supervisor

or identified during previous research 61

Inclusion of additional articles, reports

APPENDIX A. APPENDIX

Table A.1: Articles used for in depth comparison of the detection of rib fractures:

Paper Comparison

M. A.Weber, R. A. Risdon, A. C. O., M. Malone,

and N. J. Sebire (2009). Doi: 10.1016/j.forsciint.2009.04.015 CR vs. autopsy T. R. Sanchez, A. D. Grasparil, R. Chaudhari, K. P. Coulter,

and S. L. Wootton-Gorges (2018). Doi: 10.1097/PEC.0000000000000608 CR/CT vs. follow-up CR S. L. Wootton-Gorges, R. Stein-Wexler, J. W. Walton, A. J. Rosas,

K. P. Coulter, and K. K. Rogers (2008). Doi: 10.1016/j.chiabu.2007.06.011 CR vs. CT T. S. Hong, J. A. Reyes, R. Moineddin, D. A. Chiasson,

W. E. Berdon, and P. S. Babyn (2011). Doi: 10.1007/s00247-010-1953-7 CR/ CT vs. autopsy S. C. Shelmerdine, D. Langan, J. C. Hutchinson, M. Hickson,

K. Pawley, J. Suich, L. Palm, N. J. Sebire, A. Wade, and O. J. Arthurs (2018). Doi: 10.1016/S2352-4642(18)30274-8

CR/ CT vs. autopsy

Table A.2: Articles used for in depth comparison of the detection of skull fractures:

Paper Comparison

P. A. Culotta, J. E. Crowe, Q. Tran, J. Y. Jones, A. R. Mehollin-Ray, H. B. Tran, M. Donaruma-Kwoh, C. T. Dodge, E. A. Camp, and A. T. Cruz (2017). Doi: 10.1007/s00247-016-3707-7

CR vs. (3D)CT S. R. Sharp, S. M. Patel, R. E. Brown, and C. Landes

(2018). Doi: 10.1016/j.crad.2017.11.027 CR vs. (3D)CT S. P. Prabhu, A. W. Newton, J. M. Perez-Rossello,

and P. K. Kleinman (2013). Doi: 10.1007/s00247-012-2546-4 2DCT vs. 3DCT

Table A.3: Articles used for in depth comparison of the detection of CMLs:

Paper Comparison

A. Tsai, S. A. Connolly, K. Ecklund, P. R. Johnston, and P. K. Kleinman (2019). Doi: 10.1007/s00247-018-4329-z

1-view CR vs. multiview & follow-up CR