Pediatr Radiol (2014) 44:1224–1229

DOI 10.1007/s00247-014-2995-z

ORIGINAL ARTICLE

Healing patterns of clavicular birth injuries as a guide

to fracture dating in cases of possible infant abuse
Michele M. Walters & Peter W. Forbes & Carlo Buonomo &
Paul K. Kleinman

Received: 4 December 2013 / Revised: 17 February 2014 / Accepted: 30 March 2014 / Published online: 29 April 2014
# Springer-Verlag Berlin Heidelberg 2014

Abstract infant clavicular fracture healing can provide an evidence base

Background Dating fractures is critical in cases of suspected that may be applicable in cases of suspected infant abuse.
infant abuse. There are little scientific data to guide radiolo-
gists, and dating is generally based on personal experience and Keywords Radiography . Fracture dating . Fracture healing .
conventional wisdom. Child abuse . Non-accidental trauma . Neonate . Clavicle
Objective Since birth-related clavicular fractures are not
immobilized and their age is known, we propose that an
assessment of these injuries may serve as a guide for dating Introduction
inflicted fractures in young infants, acknowledging that pat-
terns observed in the clavicle may not be entirely generaliz- Accurate dating of fractures is desirable in cases of suspected
able to other bones injured in the setting of abuse. infant abuse. Determination of fracture age may support or
Materials and methods One hundred thirty-one radiographs of call into question the veracity of the history provided to
presumed birth-related clavicular fractures in infants between 0 explain a particular injury [1–3]. Multiple fractures of differ-
and 3 months of age were reviewed by two pediatric radiolo- ent ages are often seen in the setting of inflicted injury, and
gists with 30 and 15 years’ experience. Readers were asked to reliable fracture dating may be critical to determine if this
evaluate images based on several parameters of fracture pattern is present [1, 4, 5]. Accurate fracture dating can aid
healing, with a focus on subperiosteal new bone formation in the identification and exclusion of potential perpetrators in
(SPNBF) and callus formation. SPNBF and callus were each the setting of abuse [1]. There are little scientific data available
evaluated with regard to presence, thickness and character. to guide radiologists, and fracture dating is often based on
Responses were correlated with known fracture ages. personal experience and conventional wisdom [1, 6].
Results SPNBF was rarely seen in fractures less than 7 days Inflicted skeletal injuries are more common in younger age
old and was most often present by 10 days. Callus formation groups and most frequent in infants [7–10]. Previous works
was rarely seen in fractures less than 9 days old and was most have focused on patterns of fracture healing in older children,
often present by 15 days. SPNBF thickness increased with rather than the more relevant infant population [11–13]. The
fracture age and the character of SPNBF evolved from single- timetable of fracture healing is known to be quite different in
layered to solid/multilayered. Callus thickness decreased with these age groups. Some studies have examined healing pat-
fracture age and callus matrix evolved from soft to intermedi- terns in fractures that are immobilized [11–15]. This treatment
ate to hard in character. renders inferences drawn from these fractures less generaliz-
Conclusion There is an evolution in clavicular fracture able to cases of child abuse, since inflicted injuries are often
healing in young infants that follows a predictable pattern. initially untreated due to delay in diagnosis. In some studies,
These findings afford the prospect that predictable patterns of fractures were evaluated through cast material, a clear limita-
tion to detailed evaluation of healing patterns [11–15]. Other
studies have focused exclusively on later stages of fracture
M. M. Walters (*) : P. W. Forbes : C. Buonomo : P. K. Kleinman
healing such as callus formation, excluding earlier
Department of Radiology, Boston Children’s Hospital,
300 Longwood Ave., 02115 Boston, MA, USA subperiosteal new bone formation (SPNBF) from assessments
e-mail: [email protected] [11]. A reliable timetable outlining patterns of fracture healing
Pediatr Radiol (2014) 44:1224–1229 1225

should ideally include expected patterns of development of to 2009 was performed. Cases were identified employing an
SPNBF as well as callus. electronic text search for radiology reports where the follow-
The periosteum consists of two layers, an outer fibrous ing phrases appeared: clavicle fracture, birth injury, shoulder
layer and an inner osteogenic layer adjacent to the bony dystocia or large for gestational age. Injuries that were con-
cortex. In the setting of fracture, the periosteum is stripped sidered to be birth related were identified through a review of
away from the underlying cortical bone, and contained hem- radiology reports in conjunction with the clinical notes from
orrhage dissects along the bone beneath the osteogenic layer the newborn nursery, intensive care unit, primary care visits
of the periosteum. Hemorrhage is ultimately replaced by and emergency department clinical notes. Infants were ex-
vascular fibrous tissue and new bone [16]. Radiographically, cluded when relation to the time of birth could not be
subperiosteal new bone may appear as a hazy cortical margin established with a reasonable degree of certainty. Premature
or a thin layer of bone separated from the original cortex by a infants and those with any significant health issue such as
discrete lucent interval [17]. This new bone increases in congenital heart disease or sepsis were excluded, as it was felt
thickness with time and may evolve to demonstrate a that the stress associated with these conditions might affect
lamellated or multilayered configuration. observed patterns of fracture healing. Infants with potentially
Callus develops around the ends of a fractured bone by confounding factors such as skeletal dysplasia or
cellular organization within the fracture hematoma, and ulti- pseudoarthrosis were also eliminated from consideration. This
mately unites the fracture fragments. Fibrovascular tissue electronic search and medical record review yielded 142
replaces the hematoma with collagen fibers and matrix ele- cases. Following preliminary evaluation of images, 11 of these
ments, which become mineralized and form the woven bone 142 cases were eliminated. In four, image quality was felt to
of the primary callus. This process begins near the fracture be suboptimal for detailed fracture analysis. In three others, a
margins and proceeds centrifugally away from the injury site. fracture could not be confirmed. Concomitant infection was
Cartilage formation occurs, predominantly at the periphery of present in three eliminated cases. Of these, two patients were
the callus, and conversion to bone occurs through endochon- septic and concern was raised for osteomyelitis involving the
dral ossification. Radiographically, callus is initially visible as clavicle based on clinical and imaging findings. The third had
faint calcification and referred to as soft, immature or fluffy. a soft-tissue infection in the supraclavicular region, with a
As ossification progresses and callus thickens and develops a large, organized collection demonstrated on ultrasound. In
trabecular matrix, it is referred to as hard or mature [16]. one patient, accidental trauma occurred at home in the neona-
With further healing, the SPNBF blends with the callus tal period shortly after discharge from the hospital. Thus, it
forming a firm bony union across the fracture site. was not felt that clavicular injury could be tied to the date and
The birth-related clavicular fracture is an appealing surro- time of birth with an acceptable level of certainty. Following
gate for inflicted injury because fracture age at the time of preliminary image review, a cohort of 131 cases of presumed
imaging is known with reasonable certainty. Additionally, this birth-related mid shaft clavicle fractures was compiled. In this
model affords the study of the patterns of fracture healing in cohort, some infants were imaged on more than one occasion
young infants, the population at greatest risk for a poor out- separated by a minimum of 24 h. Ten infants were imaged
come [18]. Clavicular morphology is similar to that of other twice, three infants were imaged three times, and one infant
tubular bones, and therefore healing patterns may be general- was imaged on six separate occasions.
izable to fractures involving the long bones and ribs in abused
infants. Also, infant clavicular fractures are generally not Radiographic analysis
immobilized [19], and therefore the spectrum of fracture
healing may be similar to that of abusive fractures in which All clavicle fractures were evaluated by two pediatric radiolo-
treatment is delayed. The purpose of this study was to define gists (C.B. and P.K.K.) with 15 and 30 years’ experience at
predictable patterns of fracture healing in infants through the independent reading periods, respectively. These sessions were
assessment of a large sample of birth-related clavicular inju- conducted with a facilitator, using a questionnaire that was
ries to establish an evidence base that can be applied in the completed for each case as readers analyzed specific parameters
setting of suspected infant abuse. of fracture healing. Readers were blinded to infant/fracture age.
Primary focus was placed on the evaluation of SPNBF and
callus formation, as these parameters of healing were felt to be
Materials and methods radiographically distinguishable and important in discerning
age of relatively more recent fractures often seen on presen-
Inclusion criteria tation in the setting of abuse. Readers were initially trained in
the assessment of these parameters of healing, examining
A retrospective study of clavicle fractures in infants (0– fractures in various stages of healing with the help of a
3 months) imaged at two tertiary care institutions from 1997 facilitator. Cases evaluated during training were not included
1226 Pediatr Radiol (2014) 44:1224–1229

Fig. 1 Radiographs of the

clavicle in two different patients
show change in subperiosteal new
bone formation (SPNBF)
thickness and character with
fracture age. a Single-layered
SPNBF in a 10-day-old infant
(arrows). b Solid/multilayered
SPNBF in a 13-day-old infant
(arrows)

in the study group. SPNBF was described as new bone callus formation was collapsed to a two-point scale: not
paralleling the original cortex of the bone with a linear con- present and present.
figuration, and characterized as single-layered or solid/ Readers were asked to record fracture orientation when
multilayered (Fig. 1). Callus formation was described as min- possible (vertical, oblique or comminuted). Fracture margins
eralization first evident as amorphous opacity near the cortical were assessed and classified as very indistinct, somewhat indis-
margins of the fracture, and subsequently progressing centrif- tinct, somewhat defined or well defined. Fracture displacement
ugally away from the injury site with a more spherical con- was evaluated in terms of percent clavicle shaft width and
figuration. Callus was characterized as soft, intermediate or designated as non-displaced, 0–50%, 50–100% or >100%
hard in character, with soft callus having a fluffy, cloud-like displaced relative to shaft width. Fracture angulation was de-
appearance, hard callus demonstrating a solid trabecular ma- scribed as not present or mild, moderate or severe angulation.
trix and intermediate callus falling between the two (Fig. 2).
SPNBF was evaluated with regard to presence, thickness Statistical methods
and character. Presence of SPNBF was assessed on a four-
point scale, with responses including not present, probably not Data were entered using Excel and analyzed with SAS version
present, probably present and present. A designation of “pres- 9.3 (SAS Institute, Inc., Cary, NC). The agreement between
ent but indistinguishable from callus” was also available to the two readers concerning the presence or absence of SPNBF
readers. When this was selected, readers were instructed to and callus was compared using two-by-two tables and com-
skip detailed analysis of SPNBF and analyze fractures in puting kappa statistics. ANOVA, t-tests and Kruskal-Wallis
terms of callus. SPNBF thickness was evaluated relative to nonparametric tests were used to examine the relationship
clavicle shaft width, with choices including <25%, 25–75% between fracture age (continuous) and the levels of SPBNF
and >75%. SPNBF character was classified by readers as thickness (3 levels), SPNBF character (2 levels), callus thick-
single-layered or solid/multilayered. ness (3 levels), and callus matrix (3 levels). Logistic regres-
Callus formation was evaluated in terms of presence, sions were used to model the likelihood of the presence of
thickness and matrix. Callus presence was classified on a SPNBF and callus as fracture age increases.
four-point scale designated as not present, probably not
present, probably present and present. Callus thickness
was assessed relative to clavicle shaft width, with choices Results
including <50%, 50–100% and >100%. Callus matrix was
classified based on the presence or absence of trabecular Fracture ages for the 131 cases ranged from 0 to 93 days
architecture as soft, intermediate or hard. For simplicity, (mean: 13±16 days). Each case was evaluated once by each of
the four-point scale used to assess presence of SPNBF and two independent readers, yielding 262 reads.

Fig. 2 Radiographs of the clavicle in three different patients show the evolution of callus matrix with fracture age. a Soft callus formation in a 12-day-old
infant. b Intermediate callus formation in a 16-day-old infant. c Hard callus formation in a 29-day-old infant
Pediatr Radiol (2014) 44:1224–1229 1227

Overall, reader agreement regarding presence or absence of

SPNBF was excellent (kappa=0.92). Readers agreed that
SPNBF was present in 68 cases and not present in 58 cases.
In the remaining five cases (4%), reader conclusions regarding
SPNBF were different. Fracture age for the five cases where
readers disagreed ranged from 1 to 9 days old.
For fractures between 0 and 7 days of age (n=59 cases, 118
reads), readers agreed that SPNBF was not present in 54 cases
(92%), and present in two cases (3%). In three cases (5%),
readers came to different conclusions about whether SPNBF
was present. For fractures between 8 and 9 days of age (n=7
cases, 14 reads), readers agreed that SPNBF was not present in
Fig. 4 Change in subperiosteal new bone formation (SPNBF) character
three cases (43%) and present in two cases (29%). There was
with fracture age. SPNBF evolves from single-layered to solid/multilay-
disagreement regarding the other two cases. For fractures ered with fracture age
10 days and older (n=65 cases, 130 reads), SPNBF was not
present in only one read. Callus thickness relative to clavicle shaft width was noted
SPNBF thickness relative to clavicle shaft width increased to decrease with increasing fracture age, although this finding
with fracture age (Kruskal-Wallis P<0.0001) (Fig. 3). The was less robust and not significant for each statistical test
character of SPNBF evolved with fracture age. Single- conducted (mean [SD] days: mild 38 [29]; moderate 28 [21];
layered SPNBF was associated with younger fractures than large 22 [9]; ANOVA P=0.008; Kruskal-Wallis P=0.37)
solid/multilayered SPNBF (mean [SD] days: single-layered (Fig. 5). Callus matrix evolved with fracture age. A soft matrix
7.9 [2.9]; solid/multilayered 18.4 [13.3], Wilcoxon rank-sum was noted in younger fractures, and older fractures demon-
test P<0.0001) (Fig. 4). strated a hard matrix (mean [SD] days: soft 18 [8]; intermedi-
Overall, reader agreement regarding callus presence or ate 25 [10]; hard 42 [26]; Kruskal-Wallis P<0.0001) (Fig. 6).
absence was high (kappa=0.89). Readers agreed that callus Logistic regressions were used to model the presence or
was present in 50 cases and not present in 74 cases. There was absence of SPNBF and callus formation. The SPNBF model
disagreement in seven cases (5%). estimates that SPNBF is present by day eight in 50% of cases.
For fractures between 0 and 8 days of age (n=63 cases, 126 The callus model estimates that callus is present by day 13 in
reads), readers agreed that callus was not present in 100% of 50% of cases (Fig. 7).
cases. For fractures between 9 and 14 days of age (n=26 A statistically significant correlation with fracture age was
cases, 52 reads), readers agreed that callus was not present not found for other variables assessed, including fracture
in ten cases (38%) and present in nine cases (35%). In the orientation, fracture margins, displacement and angulation.
other seven cases (27%), readers disagreed about the presence
of callus. For fractures 15 days and older (n=42 cases, 84
reads), readers agreed that callus was not present in one case Discussion
(2%) and present in 41 cases (98%).
Our data demonstrate an evolution in clavicular fracture
healing in young infants that follows a predictable pattern.

Fig. 3 Change in subperiosteal new bone formation (SPNBF) thickness

is expressed as percent clavicle shaft width as a function of fracture age. Fig. 5 Change in callus thickness is expressed as percent clavicle shaft width
SPNBF increases in thickness with fracture age as a function of fracture age. Callus decreases in thickness with fracture age
1228 Pediatr Radiol (2014) 44:1224–1229

margin sclerosis. Later stages of healing were also included

in this analysis: bony bridging, periosteal incorporation and
remodeling. All fractures were studied in casts, limiting osse-
ous detail and reducing motion that might have occurred if the
injuries were not immobilized. Yeo and Reed [11] defined
staging criteria for femoral fractures in a small population of
children from birth to 14 years old, focusing only on callus
formation. These injuries were treated, with traction followed
by the application of a cast.
In more recent work, Halliday et al. [14] assessed param-
eters of healing of long bone fractures in infants investigated
for child abuse. Readers analyzed patterns of soft-tissue swell-
Fig. 6 Change in callus matrix with fracture age. Callus matrix evolves ing, SPNBF, fracture line definition, callus formation, callus
from soft to intermediate to hard in character with fracture age character and endosteal callus formation. The degree of inter-
observer agreement among readers was also evaluated, and
SPNBF is highly unlikely in fractures less than 7 days old, and found to be only moderate for all features except SPNBF.
is most often present by 10 days. Subperiosteal new bone Since the date of injury was ascertained through confession
thickness increases with fracture age and its character changes by perpetrators or the testimony of third parties, it was difficult
from single-layered to solid/multilayered. Callus formation is to estimate fracture age with a high degree of certainty. As
highly unlikely in fractures less than 9 days old and most often with several other prior studies, fractures were immobilized
present by 15 days. Callus thickness decreases with increasing and assessments were limited by the presence of cast material.
fracture age and callus matrix evolves from soft to intermedi- In 2011, Malone et al. [15] reported a timetable of fracture
ate to hard in character. healing derived from traumatic injuries of the radius and tibia. Six
There has been little prior published scientific data on stages of fracture healing were described using a classification
dating fractures. An early study by Cumming [20] focused system based on data from combined adult and pediatric studies
on the highly relevant young infant population, studying of bone healing. The authors did not make a distinction between
newborns with birth-related clavicle, humerus and femur frac- SPNBF and callus formation. Fractures were immobilized and
tures. This small study was limited to the consideration of assessments were limited by overlying cast material. This group
timing of the first appearance of calcification at the injury site, found that the healing time of younger individuals (0–1 year) was
which varied for the different fracture sites studied. Islam et al. significantly shorter than older individuals (1–5 years), and the
[12] studied and defined staging criteria for forearm fracture mean healing time of upper extremity fractures was shorter than
healing in a large population of children between 1 and the mean for lower extremity injuries.
17 years old, excluding the young infant population. In their In recent work, Prosser et al. [13] studied six parameters of
study, a large majority of cases were in children older than fracture healing in accidental long bone fractures in a large
4 years of age. Several parameters of fracture healing were sample of children between 0 and 5 years of age. In this
assessed, including SPNBF, callus formation and fracture study, mean patient age was 4.8 years at the time of initial
radiographical assessment with relatively few patients in the
infant age group included. Some fractures underwent manip-
ulation and many were evaluated through cast material. Frac-
tures were designated as acute (<1 week), recent (8–35 days),
or old (>36 days) based on radiographical assessments.
A recent study by Sanchez et al. [21] evaluated patterns of
rib fracture healing in a small sample of infants who had
sustained inflicted injury. Initial and follow-up skeletal sur-
veys were assessed, with a focus on the appearance of the
fracture line and the evolution of callus thickness over time.
Six stages of healing were defined based on these parameters
and a timetable for fracture healing was proposed. The authors
suggested that the measurement of callus thickness specifical-
ly may provide an objective means for estimating rib fracture
age. As with earlier work by Halliday [14], date of injury is
Fig. 7 Logistic regression demonstrates the probability of presence of difficult to ascertain with certainty with this study design, and
subperiosteal new bone formation (SPNBF) and callus with fracture age thus estimation of fracture age may not be entirely accurate.
Pediatr Radiol (2014) 44:1224–1229 1229

In our study we strived to overcome some of the References

shortcomings of these earlier reports, but the work has
some limitations, particularly with regard to the use of 1. O’Connor JF, Cohen J (1998) Dating fractures. In: Kleinman PK (ed)
retrospective methodology. While our review of medical Diagnostic imaging of child abuse, 2nd edn. Mosby Year Book, Inc.,
records took care to identify any cases of clavicular St. Louis, pp 168–177
2. Kleinman PK, Blackbourne BD, Marks SC et al (1989) Radiologic
fractures that may have occurred after birth, we cannot
contributions to the investigation and prosecution of cases of fatal
entirely exclude that possibility. We studied clavicle and infant abuse. N Engl J Med 320:507–511
chest radiographs acquired using routine departmental 3. Chapman S (1992) The radiological dating of injuries. Arch Dis
technique with a conventional imaging system. While Child 67:1063–1065
4. Duhaime AC, Alario AJ, Lewander WJ et al (1992) Head injury in
this approach may have offered less detail than Ameri-
very young children: mechanisms, injury types, and ophthalmologic
can College of Radiology (ACR) standardized skeletal findings in 100 hospitalized patients younger than 2 years of age.
survey images [22], we chose only those radiographs Pediatrics 90:179–185
that were deemed of sufficient technical quality to as- 5. Leventhal JM, Thomas SA, Rosenfield NS et al (1993) Fractures in
young children. distinguishing child abuse from unintentional inju-
sess the features of fracture healing under study. We
ries. Am J Dis Child 147:87–92
examined the imaging features and chronology of frac- 6. Strouse PJ, Boal DKB (2013) Child abuse. In: Coley BD (ed)
ture repair in the clavicle. While these healing patterns Caffey’s pediatric diagnostic imaging, 12th edn. Elsevier Saunders,
may apply to other tubular bones injured in the setting Philadelphia, pp 1587–1598
7. Ebbin AJ, Gollub MH, Stein AM et al (1969) Battered child syn-
of abuse, their generalization to all inflicted fractures is
drome at the Los Angeles County general hospital. Am J Dis Child
limited. We focused on mid shaft clavicular fractures, 118:660–667
and our findings may not apply to healing patterns of 8. Akbarnia B, Torg JS, Kirkpatrick J et al (1974) Manifestations of the
metaphyseal and physeal injuries. Finally, the chronolo- battered-child syndrome. J Bone Joint Surg Am 56:1159–1166
9. Herndon WA (1983) Child abuse in a military population. J Pediatr
gy of fracture healing is known to evolve with age;
Orthop 3:73–76
therefore, our findings noted in early infancy may not 10. Worlock P, Stower M, Barbor P (1986) Patterns of fractures in
be applicable to older infants and children. accidental and non-accidental injury in children: a comparative study.
Our study builds on prior research in this area but is unique Br Med J 293:100–102
in some respects. We studied a large population of young 11. Yeo LI, Reed MH (1994) Staging of healing of femoral fractures in
children. Can Assoc Radiol J 45:16–19
infants, the age group most commonly affected by inflicted 12. Islam O, Soboleski D, Symons S et al (2000) Development and
injury. We believe that our approach using a specific birth- duration of radiographic signs of bone healing in children. AJR Am
related tubular bone fracture affords a greater degree of con- J Roentgenol 175:75–78
fidence of actual fracture age than studies of inflicted fractures 13. Prosser I, Lawson Z, Evans A et al (2012) A timetable for the
radiologic features of fracture healing in young children. AJR Am J
that rely upon witnesses or perpetrator confessions. Infant Roentgenol 198:1014–1020
clavicular fractures are usually not immobilized, as is common 14. Halliday KE, Broderick NJ, Somers JM et al (2011) Dating fractures
initially with inflicted injuries, and this approach offers dis- in infants. Clin Radiol 66:1049–1054
tinct advantages to studies of various accidental fractures that 15. Malone CA, Sauer NJ, Fenton TW (2011) A radiographic assessment
of pediatric fracture healing and time since injury. J Forensic Sci 56:
have been casted. We have focused on the earliest parameters 1123–1130
of fracture healing, with a detailed analysis of the patterns of 16. Kleinman PK (1998) Skeletal trauma: general considerations. In:
development of SPNBF and callus formation. Kleinman PK (ed) Diagnostic imaging of child abuse, 2nd edn.
Mosby Year Book, Inc, St. Louis, pp 8–25
17. Ogden JA (2000) Biology of repair of the immature skeleton. In:
Ogden JA (ed) Skeletal injury in the child, 3rd edn. Springer-Verlag,
New York, pp 243–268
Conclusion 18. Child Maltreatment (2011). Children’s Bureau, Administration on
Children, Youth and Families. Published 2011. Available at: http://
These findings afford the prospect that the predictable www.acf.hhs.gov/sites/default/files/cb/cm11.pdf. Accessed July 15,
patterns of infant clavicular fracture healing can provide 2012.
19. Sarwark JF (2010) Proximal humerus, scapula, and clavicle. In: Beaty
an evidence base that may be applicable to cases of JH, Kasser JR (eds) Rockwood and Wilkins’ fractures in children, 7th
suspected infant abuse. Further work is needed to deter- edn. Lippincott Williams & Wilkins, Philadelphia, pp 625–626
mine the generalizability of our results to other bones and 20. Cumming WA (1979) Neonatal skeletal fractures. birth trauma or
other injury patterns seen in the setting of abuse. Study of child abuse? Can Assoc Radiol J 30:30–33
21. Sanchez TR, Nguyen H, Palacios W et al (2013) Retrospective
the evolution of fracture healing through later infancy and evaluation and dating of non-accidental rib fractures in infants. Clin
early childhood is also needed. Radiol 68:467–471
22. ACR-SPR (1997) Practice guideline for skeletal surveys in children. The
American College of Radiology Published 1997. Revised 2011.
Available at: http://www.acr.org/~/media/ACR/Documents/PGTS/
Conflicts of interest None. guidelines/Skeletal_Surveys.pdf. Accessed 15 July 2012.
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.