Accepted Manuscript Title: How many radiographs are needed to detect angular stable head screw cut outs of the proximal humerus–a cadaver study Author: Christian Spross Bernhard Jost Stefan Rahm Sebastian Winklhofer Johannes Erhardt Emanuel Benninger PII: DOI: Reference:
S0020-1383(14)00263-0 http://dx.doi.org/doi:10.1016/j.injury.2014.05.025 JINJ 5755
To appear in:
Injury, Int. J. Care Injured
Received date: Accepted date:
9-4-2014 18-5-2014
Please cite this article as: Spross C, Jost B, Rahm S, Winklhofer S, Erhardt J, Benninger E, How many radiographs are needed to detect angular stable head screw cut outs of the proximal humerusndasha cadaver study, Injury (2014), http://dx.doi.org/10.1016/j.injury.2014.05.025 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Radiographic Detection of Proximal Humerus Screw Perforations
1
How many radiographs are needed to detect angular stable head screw cut outs of the
2
proximal humerus – a cadaver study
3 Christian Spross, MD1, Bernhard Jost, MD1, Stefan Rahm, MD2, Sebastian Winklhofer, MD3,4,
5
Johannes Erhardt, MD1, Emanuel Benninger, MD1
cr
ip t
4
6 1
Department of Orthopaedics and Traumatology, Kantonsspital St. Gallen, Switzerland
8
2
Department of Orthopaedics, University Hospital Balgrist, Zurich Switzerland
9
3
Department of Forensic Medicine and Radiology, Institute of Forensic Medicine University of
Zurich, Switzerland
11
4
M
10
an
us
7
Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, Switzerland
d
12 Correspondence to:
14
Bernhard Jost, Kantonsspital St. Gallen, Rorschacherstr. 95, 9007 St. Gallen,
15
Phone: +41 71 494 13 62 / Fax: +41 71 494 61 56, e-mail: [email protected]
Ac ce p
16
te
13
17
Disclaimer:
18
None of the other authors, or any member of their families, received any financial support related
19
to the subject of this article.
20 21
Acknowledgements:
22
We thank Frau M. Vich for the organization and help with the cadavers. Further on, we thank
23
Prof. Dr. med. M. Thali for offering his facilities for the radiographic investigations.
24 1
Page 1 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
24
How many radiographs are needed to detect angular stable head screw cut outs of the
25
proximal humerus – a cadaver study
26 Abstract
28
Introduction: Cut out of locking head screws is the most common complication of locking plates
29
in fracture fixation of the proximal humerus with potentially disastrous consequences. Aim of the
30
study was to find the single best and combination of radiographic projections to reliably detect
31
screw cut outs.
32
Material and Methods: The locking plate was fixed to 6 cadaveric proximal humeri. Six different
33
radiographs were performed: anteriorposterior in internal (apIR), in neutral (ap0) and in 30°
34
external rotation (apER); axial in 30° (ax30) and 60° (ax60) abduction and an outlet view. Each
35
head screw (n=9) was sequentially exchanged to perforate the humeral head with the tip and all
36
radiographs were repeated for each cut out. Randomized image reading by two blinded examiners
37
for cut out was done for single projection and combinations.
38
Results: Interrater agreement was 0.72 - 0.93. Best single projection was ax30 (sensitivity 76%)
39
and the worst was the outlet view (sens. 17%). Standard combination of apIR/outlet reached a
40
sens. of 54%. The best combination of two was: apER/ax30 (90% sens.), of three:
41
apIR/apER/ax30 (96% sens) and of four: apIR/ap0/apER/ax30 (100% sens.).
42
Conclusion: Standard radiographs (ap/outlet), especially in internal rotation, may miss nearly half
43
of screw cut outs. Single best radiographic projection was an axial view with 30° abduction. To
44
account for all cut outs and correct screw position a combination of 4 projections was needed.
45
These simple and feasible intraoperative and postoperative radiographs help to detect screw
46
perforations of the locking plate reliably.
Ac ce p
te
d
M
an
us
cr
ip t
27
47 2
Page 2 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
48
Key words: proximal humerus fracture, locking plate, cut out, complications, locking head screw,
49
radiographic projections
50 Level of evidence: I (Study of Diagnostic Test)
ip t
51
Ac ce p
te
d
M
an
us
cr
52
3
Page 3 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
Introduction
53
The PHILOS® plate (Proximal Humerus Internal Locking System, Depuy Synthes, Inc.; West
54
Chester, PA, USA) is a commonly used and reported implant for open reduction and internal
55
fixation of proximal humerus fractures.1,3,5,8-10,13-15,17 In addition to the anatomically contoured
56
shape and the low profile, the divergent head screw directions and their angular stability are
57
supposed to provide better fixation in the head fragment especially for osteoporotic bone.
58
Biomechanical advantages for stability compared to conventional plates have been shown.2
59
However, with the increasing clinical use of these implants, new kinds of complications are seen,
60
which were previously unknown in the treatment of proximal humerus fractures and specific to
61
the angular stability of the implant. Clinical investigations further confirmed perforations of the
62
angular stable head screws to be the most common complications with revision surgery related to
63
this kind of plates.1,10,14-17 There are two main causes of this complication. First, the divergent
64
head screw angles in combination with the sphericity of the humeral head makes the exact
65
localization of the screw tip in the head fragment difficult during intraoperative fluoroscopy and
66
can result in primary screw cut out. Second, because of the often-impaired medial support and the
67
fragile blood supply of the head fragment, fractures of the proximal humerus are prone to
68
secondary dislocation and avascular necrosis.5,10,15-17 When occurring, the angular stability of the
69
head screws does not allow any slip back and the screw tips slowly protrude through the head
70
fragment into the joint, also known as secondary screw cut out. Once in the joint, the screw tip
71
can cause damage to the glenoid. Pain seems to be a unreliable symptom for screw cut out, as
72
clinical cases are described with very late detection of head screw cut outs and severe secondary
73
destruction of the glenoid bone stock.5 These reasons make an early reliable radiographic
74
detection of potential head screw cut out crucial.
Ac ce p
te
d
M
an
us
cr
ip t
52
4
Page 4 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
There is a lack of information in the literature about the detailed radiographic investigations,
76
which are needed to rule out head screw cut outs of the proximal humerus during surgery and
77
postoperative follow-up. After a review of clinical studies published on this topic, none was
78
found to describe precisely the used intra-operative projections of fluoroscopy. The most
79
commonly used radiographic investigations for follow-up contained a standard anterior posterior
80
and a Neer view,1,8,9,15,17 sometimes with an axillary projection.3,13 Only a small number of
81
studies considered further radiographic investigations.5,10,14 Therefore, it was the aim of this
82
cadaver study to find simple and feasible radiographic projections (single best and the best
83
combination) to reliably detect head screw cut out of a frequently used locking plate in daily
84
routine.
M
an
us
cr
ip t
75
Ac ce p
te
d
85
5
Page 5 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
Materials and Methods
86
A pilot study was performed in order to find the best possible radiographic projection to detect
87
humeral head screw cut out. A locking plate of the Proximal Humerus Internal Locking System
88
(PHILOS®: Depuy Synthes, Inc.; West Chester, PA, USA) was fixed on an anatomical right
89
composite humerus bone (5010 Humerus; Synbone, Malans, Switzerland). All head screws
90
perforated the cartilage of the humeral head with 1mm of their tips. Different radiographic
91
projections were performed with fluoroscopy and the protruding screw tips were counted. The
92
standard radiographs which are the axial view in 60-90° (ax60) of abduction and neutral rotation
93
as well as the Outlet- or Neer-view and the anterior posterior views in internal (apIR) and neutral
94
rotation (ap0) were not able to detect all screw cut outs. Therefore, an additional axial view was
95
added to this series with the arm in 30° of abduction and neutral rotation which showed more
96
screw tips from the inferior humeral “hemisphere” compared to the standard axial view in 90° of
97
abduction. A second projection with a high detection rate was an anterior posterior view in 30° of
98
external rotation (apER) from the central ray, which has been described in another study before.11
99
This view especially depicted the posterior humeral “hemisphere”. Both of these additional added
Ac ce p
te
d
M
an
us
cr
ip t
85
100
projections seemed to be clinically feasible intraoperatively and during later postoperative
101
radiographic follow-up as they would not cause high stress on the fracture fixation and should not
102
be painful.
103
Preparation of specimens
104
This study was performed with 6 healthy shoulders of 3 whole-body human cadavers (1 male, 2
105
females; mean age 84 (range: 83-85) years). To simulate a realistic intraoperative situation, the
106
bodies were put in a beach chair position. The proximal part of the humerus was dissected
107
through a standard deltopectoral approach. The locking plate was fixed temporary by K-wires
108
with the upper margin 8mm distal to the tip of the greater tuberosity using the manufacturers’ 6
Page 6 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
provided aiming device. The plate was placed 5mm lateral to the bicipital groove in all shoulders.
110
A non-locking screw was applied to the most proximal hole of the shaft. Before setting the head
111
screws, the subscapularis muscle was tenotomized close to the lesser tuberosity and the anterior
112
capsule was vertically incised to gain full sight into the joint and on the humeral head. Drilling
113
was stopped subchondrally without perforating the head, which was visually controlled. The
114
length of the drill hole was measured and all screws were chosen 4mm shorter.15
115
Radiographic investigation
116
A series of the following six different fluoroscopic projections was made (Figure 1, A/B/C/D).
117
For the anteroposterior projections the c-arm was positioned in anteroposterior direction
118
(perpendicular to the glenoid) with the shoulder first in internal rotation and the forearm on the
119
belly simulating the “arm in a sling” position (apIR, Fig. 1B). Then, without moving the c-arm,
120
the forearm was held exactly into the central ray (laser assisted) for a standard anteroposterior
121
view (ap0, Fig. 1C) and subsequently rotated in a 30° external rotation position, which was
122
confirmed with a goniometer in reference to the central ray, for the ap30 view (Fig. 1D). The c-
123
arm was then moved lateral for the standard outlet view according Liotard et al.7 and finally fixed
124
in vertical position for the two other axial views (Fig. 1E). The arm was first held in 30°
125
abduction and neutral rotation for the ax30 view (Fig. 1F) and then in 60° of abduction and
126
neutral rotation for the second axial projection (ax60, Fig. 1G), which we used as a standard axial
127
view that seems clinically feasible for early postoperative radiographic follow-up examinations
128
with limited abduction.
129
This series of six images was performed for each shoulder with correct length of the head screws.
130
Then, each head screw (n=9) was replaced turn by turn with a screw just perforating the cartilage
131
of the humeral head with 1mm of its tip. This was visually controlled (Figure 2). All six
132
radiographic projections were repeated for every perforating head screw giving a total of 60
Ac ce p
te
d
M
an
us
cr
ip t
109
7
Page 7 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
radiographs for each shoulder, which were saved and coded for projection and the perforated
134
head screw position.
135
Analysis of radiographs
136
For radiographic analysis, the position of all 9 head screws was numbered (Figure 3). For reading
137
the radiographs, the OsiriX DICOM viewer (version 5.0.2. 32-bit; Pixmeo SARL, 266 Rue de
138
Bernex, CH-1233 Bernex, Switzerland) was used. In a first analysis, all 360 radiographs were
139
independently and randomly reviewed for head screw cut outs (yes/no) by two readers (E.B. and
140
C.S.), blinded for the number of shoulder, the type of projection and the eventual cut out screw
141
position. Interrater reliability was calculated and, in case of disagreement, a first consensus
142
reading was performed, where both readers had to decide together, still blinded for the type of
143
projection, whether the screw was in or out. This consensus reading was needed for further
144
calculations of the sensitivity of each single projection and of all possible combinations.
145
In addition, the screw tips were defined, due to their fixed angular direction, to end either in an
146
anterior/posterior and/or superior/inferior “hemisphere” of the humeral head (figure 3). The
147
ascending screw ticketed with number 7 was defined to end in the superior half, but was not
148
classified for anterior/posterior position due to its vertical and centered position in the plate. The
149
3 different anteroposterior and the 2 axial views were compared to each other regarding best
150
projection of screw tips in the anterior/posterior and superior/inferior hemisphere respectively.
151
The same two observers independently reviewed the best calculated combinations of projections
152
again 3 months after the first reading, to minimize recall bias. This time, the readers had to decide
153
if there is a cut out (yes/no) and which screw by given number (see figure 3) exactly protrudes
154
into the joint (identification possible: yes/no, if yes: number of screw). The interrater agreement
155
for cut outs and screw positions were calculated for the combinations. In case of disagreement,
156
further consensus readings were performed for each combination of projections (apER/ax30;
Ac ce p
te
d
M
an
us
cr
ip t
133
8
Page 8 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
apIR/apER/ax30; apIR/ap0/apER/ax30), on which the calculations of sensitivity for screw cut
158
outs and screw positions were based.
159
Statistics
160
The interrater agreement for determining screw cut out (yes/no) in the single projections and
161
screw cut out (yes/no) and screw position in the double, triple and quadruple combinations was
162
assessed with the use of Cohan’s kappa coefficients respecting the guidelines described by Landis
163
and Koch.6 Sensitivity was calculated using Chi-squared tables of contingency for all single
164
projections, all possible combinations of them and for the particularly investigated combinations
165
from the second reading. In the study design, the standard of reference had to be distinct with
166
only one possible screw cut out. This resulted in no “false positive” options, thus specificity was
167
not calculated.
168
For comparison of the different anteroposterior and axial views concerning the projection of the
169
defined humeral head hemispheres, a Chi-Square test was used with a level of significance at p
S0020-1383(14)00263-0 http://dx.doi.org/doi:10.1016/j.injury.2014.05.025 JINJ 5755
To appear in:
Injury, Int. J. Care Injured
Received date: Accepted date:
9-4-2014 18-5-2014
Please cite this article as: Spross C, Jost B, Rahm S, Winklhofer S, Erhardt J, Benninger E, How many radiographs are needed to detect angular stable head screw cut outs of the proximal humerusndasha cadaver study, Injury (2014), http://dx.doi.org/10.1016/j.injury.2014.05.025 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Radiographic Detection of Proximal Humerus Screw Perforations
1
How many radiographs are needed to detect angular stable head screw cut outs of the
2
proximal humerus – a cadaver study
3 Christian Spross, MD1, Bernhard Jost, MD1, Stefan Rahm, MD2, Sebastian Winklhofer, MD3,4,
5
Johannes Erhardt, MD1, Emanuel Benninger, MD1
cr
ip t
4
6 1
Department of Orthopaedics and Traumatology, Kantonsspital St. Gallen, Switzerland
8
2
Department of Orthopaedics, University Hospital Balgrist, Zurich Switzerland
9
3
Department of Forensic Medicine and Radiology, Institute of Forensic Medicine University of
Zurich, Switzerland
11
4
M
10
an
us
7
Institute of Diagnostic and Interventional Radiology, University Hospital Zurich, Switzerland
d
12 Correspondence to:
14
Bernhard Jost, Kantonsspital St. Gallen, Rorschacherstr. 95, 9007 St. Gallen,
15
Phone: +41 71 494 13 62 / Fax: +41 71 494 61 56, e-mail: [email protected]
Ac ce p
16
te
13
17
Disclaimer:
18
None of the other authors, or any member of their families, received any financial support related
19
to the subject of this article.
20 21
Acknowledgements:
22
We thank Frau M. Vich for the organization and help with the cadavers. Further on, we thank
23
Prof. Dr. med. M. Thali for offering his facilities for the radiographic investigations.
24 1
Page 1 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
24
How many radiographs are needed to detect angular stable head screw cut outs of the
25
proximal humerus – a cadaver study
26 Abstract
28
Introduction: Cut out of locking head screws is the most common complication of locking plates
29
in fracture fixation of the proximal humerus with potentially disastrous consequences. Aim of the
30
study was to find the single best and combination of radiographic projections to reliably detect
31
screw cut outs.
32
Material and Methods: The locking plate was fixed to 6 cadaveric proximal humeri. Six different
33
radiographs were performed: anteriorposterior in internal (apIR), in neutral (ap0) and in 30°
34
external rotation (apER); axial in 30° (ax30) and 60° (ax60) abduction and an outlet view. Each
35
head screw (n=9) was sequentially exchanged to perforate the humeral head with the tip and all
36
radiographs were repeated for each cut out. Randomized image reading by two blinded examiners
37
for cut out was done for single projection and combinations.
38
Results: Interrater agreement was 0.72 - 0.93. Best single projection was ax30 (sensitivity 76%)
39
and the worst was the outlet view (sens. 17%). Standard combination of apIR/outlet reached a
40
sens. of 54%. The best combination of two was: apER/ax30 (90% sens.), of three:
41
apIR/apER/ax30 (96% sens) and of four: apIR/ap0/apER/ax30 (100% sens.).
42
Conclusion: Standard radiographs (ap/outlet), especially in internal rotation, may miss nearly half
43
of screw cut outs. Single best radiographic projection was an axial view with 30° abduction. To
44
account for all cut outs and correct screw position a combination of 4 projections was needed.
45
These simple and feasible intraoperative and postoperative radiographs help to detect screw
46
perforations of the locking plate reliably.
Ac ce p
te
d
M
an
us
cr
ip t
27
47 2
Page 2 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
48
Key words: proximal humerus fracture, locking plate, cut out, complications, locking head screw,
49
radiographic projections
50 Level of evidence: I (Study of Diagnostic Test)
ip t
51
Ac ce p
te
d
M
an
us
cr
52
3
Page 3 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
Introduction
53
The PHILOS® plate (Proximal Humerus Internal Locking System, Depuy Synthes, Inc.; West
54
Chester, PA, USA) is a commonly used and reported implant for open reduction and internal
55
fixation of proximal humerus fractures.1,3,5,8-10,13-15,17 In addition to the anatomically contoured
56
shape and the low profile, the divergent head screw directions and their angular stability are
57
supposed to provide better fixation in the head fragment especially for osteoporotic bone.
58
Biomechanical advantages for stability compared to conventional plates have been shown.2
59
However, with the increasing clinical use of these implants, new kinds of complications are seen,
60
which were previously unknown in the treatment of proximal humerus fractures and specific to
61
the angular stability of the implant. Clinical investigations further confirmed perforations of the
62
angular stable head screws to be the most common complications with revision surgery related to
63
this kind of plates.1,10,14-17 There are two main causes of this complication. First, the divergent
64
head screw angles in combination with the sphericity of the humeral head makes the exact
65
localization of the screw tip in the head fragment difficult during intraoperative fluoroscopy and
66
can result in primary screw cut out. Second, because of the often-impaired medial support and the
67
fragile blood supply of the head fragment, fractures of the proximal humerus are prone to
68
secondary dislocation and avascular necrosis.5,10,15-17 When occurring, the angular stability of the
69
head screws does not allow any slip back and the screw tips slowly protrude through the head
70
fragment into the joint, also known as secondary screw cut out. Once in the joint, the screw tip
71
can cause damage to the glenoid. Pain seems to be a unreliable symptom for screw cut out, as
72
clinical cases are described with very late detection of head screw cut outs and severe secondary
73
destruction of the glenoid bone stock.5 These reasons make an early reliable radiographic
74
detection of potential head screw cut out crucial.
Ac ce p
te
d
M
an
us
cr
ip t
52
4
Page 4 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
There is a lack of information in the literature about the detailed radiographic investigations,
76
which are needed to rule out head screw cut outs of the proximal humerus during surgery and
77
postoperative follow-up. After a review of clinical studies published on this topic, none was
78
found to describe precisely the used intra-operative projections of fluoroscopy. The most
79
commonly used radiographic investigations for follow-up contained a standard anterior posterior
80
and a Neer view,1,8,9,15,17 sometimes with an axillary projection.3,13 Only a small number of
81
studies considered further radiographic investigations.5,10,14 Therefore, it was the aim of this
82
cadaver study to find simple and feasible radiographic projections (single best and the best
83
combination) to reliably detect head screw cut out of a frequently used locking plate in daily
84
routine.
M
an
us
cr
ip t
75
Ac ce p
te
d
85
5
Page 5 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
Materials and Methods
86
A pilot study was performed in order to find the best possible radiographic projection to detect
87
humeral head screw cut out. A locking plate of the Proximal Humerus Internal Locking System
88
(PHILOS®: Depuy Synthes, Inc.; West Chester, PA, USA) was fixed on an anatomical right
89
composite humerus bone (5010 Humerus; Synbone, Malans, Switzerland). All head screws
90
perforated the cartilage of the humeral head with 1mm of their tips. Different radiographic
91
projections were performed with fluoroscopy and the protruding screw tips were counted. The
92
standard radiographs which are the axial view in 60-90° (ax60) of abduction and neutral rotation
93
as well as the Outlet- or Neer-view and the anterior posterior views in internal (apIR) and neutral
94
rotation (ap0) were not able to detect all screw cut outs. Therefore, an additional axial view was
95
added to this series with the arm in 30° of abduction and neutral rotation which showed more
96
screw tips from the inferior humeral “hemisphere” compared to the standard axial view in 90° of
97
abduction. A second projection with a high detection rate was an anterior posterior view in 30° of
98
external rotation (apER) from the central ray, which has been described in another study before.11
99
This view especially depicted the posterior humeral “hemisphere”. Both of these additional added
Ac ce p
te
d
M
an
us
cr
ip t
85
100
projections seemed to be clinically feasible intraoperatively and during later postoperative
101
radiographic follow-up as they would not cause high stress on the fracture fixation and should not
102
be painful.
103
Preparation of specimens
104
This study was performed with 6 healthy shoulders of 3 whole-body human cadavers (1 male, 2
105
females; mean age 84 (range: 83-85) years). To simulate a realistic intraoperative situation, the
106
bodies were put in a beach chair position. The proximal part of the humerus was dissected
107
through a standard deltopectoral approach. The locking plate was fixed temporary by K-wires
108
with the upper margin 8mm distal to the tip of the greater tuberosity using the manufacturers’ 6
Page 6 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
provided aiming device. The plate was placed 5mm lateral to the bicipital groove in all shoulders.
110
A non-locking screw was applied to the most proximal hole of the shaft. Before setting the head
111
screws, the subscapularis muscle was tenotomized close to the lesser tuberosity and the anterior
112
capsule was vertically incised to gain full sight into the joint and on the humeral head. Drilling
113
was stopped subchondrally without perforating the head, which was visually controlled. The
114
length of the drill hole was measured and all screws were chosen 4mm shorter.15
115
Radiographic investigation
116
A series of the following six different fluoroscopic projections was made (Figure 1, A/B/C/D).
117
For the anteroposterior projections the c-arm was positioned in anteroposterior direction
118
(perpendicular to the glenoid) with the shoulder first in internal rotation and the forearm on the
119
belly simulating the “arm in a sling” position (apIR, Fig. 1B). Then, without moving the c-arm,
120
the forearm was held exactly into the central ray (laser assisted) for a standard anteroposterior
121
view (ap0, Fig. 1C) and subsequently rotated in a 30° external rotation position, which was
122
confirmed with a goniometer in reference to the central ray, for the ap30 view (Fig. 1D). The c-
123
arm was then moved lateral for the standard outlet view according Liotard et al.7 and finally fixed
124
in vertical position for the two other axial views (Fig. 1E). The arm was first held in 30°
125
abduction and neutral rotation for the ax30 view (Fig. 1F) and then in 60° of abduction and
126
neutral rotation for the second axial projection (ax60, Fig. 1G), which we used as a standard axial
127
view that seems clinically feasible for early postoperative radiographic follow-up examinations
128
with limited abduction.
129
This series of six images was performed for each shoulder with correct length of the head screws.
130
Then, each head screw (n=9) was replaced turn by turn with a screw just perforating the cartilage
131
of the humeral head with 1mm of its tip. This was visually controlled (Figure 2). All six
132
radiographic projections were repeated for every perforating head screw giving a total of 60
Ac ce p
te
d
M
an
us
cr
ip t
109
7
Page 7 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
radiographs for each shoulder, which were saved and coded for projection and the perforated
134
head screw position.
135
Analysis of radiographs
136
For radiographic analysis, the position of all 9 head screws was numbered (Figure 3). For reading
137
the radiographs, the OsiriX DICOM viewer (version 5.0.2. 32-bit; Pixmeo SARL, 266 Rue de
138
Bernex, CH-1233 Bernex, Switzerland) was used. In a first analysis, all 360 radiographs were
139
independently and randomly reviewed for head screw cut outs (yes/no) by two readers (E.B. and
140
C.S.), blinded for the number of shoulder, the type of projection and the eventual cut out screw
141
position. Interrater reliability was calculated and, in case of disagreement, a first consensus
142
reading was performed, where both readers had to decide together, still blinded for the type of
143
projection, whether the screw was in or out. This consensus reading was needed for further
144
calculations of the sensitivity of each single projection and of all possible combinations.
145
In addition, the screw tips were defined, due to their fixed angular direction, to end either in an
146
anterior/posterior and/or superior/inferior “hemisphere” of the humeral head (figure 3). The
147
ascending screw ticketed with number 7 was defined to end in the superior half, but was not
148
classified for anterior/posterior position due to its vertical and centered position in the plate. The
149
3 different anteroposterior and the 2 axial views were compared to each other regarding best
150
projection of screw tips in the anterior/posterior and superior/inferior hemisphere respectively.
151
The same two observers independently reviewed the best calculated combinations of projections
152
again 3 months after the first reading, to minimize recall bias. This time, the readers had to decide
153
if there is a cut out (yes/no) and which screw by given number (see figure 3) exactly protrudes
154
into the joint (identification possible: yes/no, if yes: number of screw). The interrater agreement
155
for cut outs and screw positions were calculated for the combinations. In case of disagreement,
156
further consensus readings were performed for each combination of projections (apER/ax30;
Ac ce p
te
d
M
an
us
cr
ip t
133
8
Page 8 of 27
Radiographic Detection of Proximal Humerus Screw Perforations
apIR/apER/ax30; apIR/ap0/apER/ax30), on which the calculations of sensitivity for screw cut
158
outs and screw positions were based.
159
Statistics
160
The interrater agreement for determining screw cut out (yes/no) in the single projections and
161
screw cut out (yes/no) and screw position in the double, triple and quadruple combinations was
162
assessed with the use of Cohan’s kappa coefficients respecting the guidelines described by Landis
163
and Koch.6 Sensitivity was calculated using Chi-squared tables of contingency for all single
164
projections, all possible combinations of them and for the particularly investigated combinations
165
from the second reading. In the study design, the standard of reference had to be distinct with
166
only one possible screw cut out. This resulted in no “false positive” options, thus specificity was
167
not calculated.
168
For comparison of the different anteroposterior and axial views concerning the projection of the
169
defined humeral head hemispheres, a Chi-Square test was used with a level of significance at p
