Biomechanical and anatomic studies of total shoulder of the results
should be improved by the use of contemporary implant designs that accurately
reproduce normal Barthroplasty implants have suggested that the durability
shoulder anatomy and thereby reproduce glenohumeral joint
kinematics1,2.
Glenoid loosening continues to be the primary reason for failure of total
shoulder
arthroplasty3-6.
Radiolucent lines at the bone-cement interface have been reported in
association with the majority of keeled glenoid components that have been
studied. For example, Lazarus et
al.7 reported
lucencies adjacent to thirty-eight of thirty-nine keeled implants that they
assessed. The rate of progressive lucencies has been reported to range from
forty-five of 146 implants to as high as eighteen of twenty-four
implants3,5,7,8.
Pegged glenoid designs were developed and introduced into the market because
of their potential for improved fixation characteristics and less bone
loss.
There have been few studies of the results of total shoulder replacement
with a pegged glenoid
implant9-13.
Radiographs have been the primary method for detecting signs of loosening.
However, the lack of a standardized radiographic technique and confounding
overlapping structures can lead to poor interobserver and intraobserver
reliability7.
Computed tomography could provide more accurate and reproducible imaging to
detect glenoid loosening and thereby become an alternative for the assessment
of glenoid implant stability.
The purpose of our study was to present the radiographic and clinical
results of total shoulder replacement with use of a four-pegged, cemented,
convex-back polyethylene glenoid implant. A new classification system for the
assessment of radiolucencies around the pegs with computed tomography was
developed and utilized to analyze the performance of these glenoid
components.
Patient Selection
Between 1999 and 2001, fifty-seven shoulders in fifty-two
consecutive patients were treated with a total shoulder arthroplasty
(Anatomical Shoulder; Zimmer, Warsaw, Indiana) either by or under the direct
supervision of the senior author (C.G.). This study was approved by the
investigational review board of our institution. All enrolled patients
consented to participate in the project after a thorough explanation of risks
and benefits. The surgical indications included destruction of the joint
surface with pain and functional limitations that were unresponsive to
conservative treatment. Exclusion criteria were postoperative infection (one
patient), incomplete follow-up (three), posttraumatic osteoarthritis with
tuberosity malunion and severe bone deformity (one), rotator cuff
insufficiency (one), and the use of a metal-backed glenoid implant (three).
This left forty-seven shoulders in forty-three patients (twenty-one men and
twenty-two women) with an average age of fifty-seven years (range, twenty-six
to eighty-one years) at the time of the index surgery. The diagnosis was
posttraumatic for nine shoulders, which had undergone an average of 0.88
previous shoulder operations; instability arthritis (osteoarthritis secondary
to instability or surgery for the treatment of instability) for seven
shoulders, which had undergone an average of 2.2 prior operations; primary
arthritis for nineteen shoulders, which had undergone an average of 0.2 prior
operations; and inflammatory arthritis for twelve shoulders, which had not
undergone any previous operations.
An Anatomical Shoulder prosthesis with an all-polyethylene glenoid
component (one of three sizes) that had a convex, roughened back surface and
four threaded pegs was used in all shoulders. The radial mismatch, defined as
the difference between the curvatures of the glenoid and the humeral head of
the total shoulder prosthesis, ranged from 0 to -4 mm. (With a noncongruent
articulation, the radius of curvature differs between the two components.)
Sixteen total shoulder replacements had a mismatch of =1 mm; twenty-four,
of 2 to 3 mm; and seven, of 4 mm. All pegs were 5 mm in diameter, and the
drill holes were 6 mm in diameter (Fig.
1). The structural features of the implant were designed
specifically to optimize fixation
strength14. All
glenoid components were fixed with low-viscosity cement (Allofix; Zimmer,
Winterthur, Switzerland) with use of a standardized
technique15.
The diagnosis, limb dominance, comorbidities, gender, age, and previous
operations were determined from chart reviews and patient interviews and were
recorded for each patient. Intraoperatively recorded data included the sizes
of the humeral and glenoid implants, head-glenoid mismatch ratios, rotator
cuff integrity, and complications.
Clinical Evaluation
Preoperative and postoperative evaluations were conducted with use of the
scoring system of Constant and
Murley16. Each
evaluation was performed by a physician who was not the operating surgeon and
who had been trained in the use of the scoring system. Abduction strength was
measured at the wrist with use of an isometric dynamometer (Isobex; Cursor,
Bern, Switzerland) with the patient standing, the arm held in 90° of
abduction in the scapular plane, the wrist fixed in pronation, and the elbow
fully extended. Three consecutive measurements, each of a three-second
duration, were made, and the average of the three measurements defined
abductor strength. Relative Constant scores were calculated with use of the
normative values obtained in Constant's original
study17. A
subjective shoulder value was determined by asking the patient to express the
value of the shoulder as a percentage of a completely normal
shoulder18.
Operative Technique
The operative technique for total shoulder replacement has been discussed
elsewhere3,5,8,
but we will describe some distinguishing features of the procedures done in
our patients. Patients underwent osteotomy of the lesser tuberosity in
continuity with the subscapularis tendon and the middle glenohumeral ligament
for exposure of the joint. The subscapularis was released by capsulotomy at
the level of the joint, mobilization of the muscle in the subscapularis fossa,
and release of adhesions from under the conjoined tendon. Glenoid version was
corrected with use of a glenoid alignment device that allowed the placement of
a guide pin at a corrected angle, predetermined with use of preoperative
computed tomography scans, and with the subsequent use of cannulated reamers.
All glenoid components were sized with a guide template and were cemented in
place. Low-viscosity cement was introduced into the peg holes, and the cement
was pressurized manually in the hole with a sponge. The process was repeated
to ensure hemostasis and optimal filling of the holes. A thin layer of cement
was also placed along the back surface of the glenoid implant before
implantation. The subscapularis and the lesser tuberosity were repaired
anatomically with transosseous nonabsorbable number-5 Ethibond sutures
(Johnson and Johnson, Norderstedt, Germany).
Radiographic Evaluation
Anteroposterior, axillary, and supraspinatus outlet radiographs were made,
under fluoroscopic guidance, preoperatively, at the first postoperative visit,
and every six months to one year thereafter until the most recent follow-up
visit. Immediate postoperative radiographs of all forty-seven shoulders were
available for analysis. Preoperative computed tomography scans were available
for forty-two shoulders. Radiographs and computed tomography scans were
acquired at the most recent follow-up visit for all forty-seven shoulders.
Images were made with the arm at the side of the body and the thumb pointing
upward. All computed tomography scans were performed on a Somatom HiQ-S
scanner (Siemens, Erlangen, Germany) with 3-mm contiguous sections from the
acromioclavicular joint to the inferior angle of the scapula.
Glenoid version was measured preoperatively and postoperatively on computed
tomography scans with use of the technique described by Friedman et
al.19,20,
and the preoperative measurements were used to quantify the correction to be
accomplished with glenoid reaming. Retroversion of 0° to 10° was
considered normal version. Glenoid erosion associated with primary arthritis
was classified with the method described by Walch et
al.21. Humeral head
subluxation with respect to the glenoid was determined on computed tomography
scans as described by Walch et al. Rotator cuff muscle quality and tendon
integrity were evaluated on computed tomography scans by a fellowship-trained
musculoskeletal radiologist. Fatty degeneration of the supraspinatus was
determined on horizontal computed tomography cuts one section (5 mm) above the
superior border of the humeral head. The infraspinatus and subscapularis
muscles were assessed, according to the technique of Goutallier et
al.22, at both the
level of the coracoid tip and the level of the infraglenoid rim. Degeneration
of the teres minor muscle was evaluated at the level of the infraglenoid
rim.
All radiographs were of adequate quality for analysis. Anteroposterior
radiographs were considered to be adequate when a clear space was visible
between the prosthetic humeral head and the glenoid component and there was
minimal overlap of the glenoid face. Axillary radiographs were considered to
be adequate when a clear space was visible between the glenoid and the
coracoid process anteriorly and the glenoid and the scapular spine
posteriorly. Radiolucent lines were scored, on a scale of 0 to 5, on
anteroposterior radiographs with use of the pegged-glenoid classification
system described by Lazarus et
al.7. Progression of
radiolucent lines was evaluated by comparing immediate postoperative
radiographs with the most recent radiographs. The amount of implant-bone
contact, or seating, was evaluated on anteroposterior and axillary radiographs
with use of the classification system proposed by Lazarus et al. Seating
reflects the amount of host subchondral bone that is directly in contact with
the back of the glenoid component.
Because there can be difficulties with visualization of radiolucencies on
radiographs, computed tomography was utilized as an additional imaging
technique. Images were directly analyzed from digitized scans with a window of
4000 and a level of 800. Three-millimeter-thick axial views from the superior
to the inferior border of the glenoid were evaluated. The established
radiolucent line score (RLL score) described by Molé et
al.8 was adapted for
computed tomography scans and to evaluate radiolucent lines around pegged
rather than keeled glenoid components. This system was originally developed to
study radiographic signs of loosening of keeled glenoid components. Our
modified system for evaluating radiolucent lines on computed tomography scans
(computed tomography lucency score) was a numeric summation of scores based on
the variable thicknesses of radiolucencies (range, 0 to 2 mm) in six different
zones along the glenoid back surface, for a total score ranging from 0 to 18
points. Zones 1 through 6 corresponded to assigned pegs or glenoid backside
areas (Fig. 2). A score of 0
was equivalent to no lucencies in any of the six zones, a score of 1 indicated
a 1-mm lucency within one zone, and a score of 2 indicated either two
lucencies measuring 1 mm each in two different zones or one lucency measuring
2 mm in one zone. A score of 18 was calculated when there was a continuous
lucency in all six zones. The largest complete radiolucent line visualized in
any zone was used to calculate the overall score.
Each radiograph and computed tomography scan was scored four times, twice
by two separate reviewers at two different time-periods. The examinations were
randomly ordered, and the reviewers were blinded with regard to the identities
of the patient and the operating surgeon and the clinical result. The
examiners were two fellowship-trained orthopaedic shoulder surgeons who were
not the operating surgeons. The most recent postoperative radiographs and
computed tomography scans were analyzed directly from digitized images on a
large-screen computer monitor at the previously stated settings. The
statistical analysis was based on the means of the individual scores provided
by each reviewer.
The degree of loosening was determined on the basis of the radiographic and
computed tomography scores. Possible loosening was defined as a radiographic
Lazarus score of =3 or a computed tomography lucency score between 6 and
12. Definite glenoid loosening was defined as a radiographic Lazarus score of
=4 or a computed tomography score of >12.
Follow-up
All patients had a radiographic assessment, which was performed at an
average of forty months (range, nineteen to fifty-five months). Forty-six
shoulders were evaluated clinically at an average of forty months (range,
nineteen to fifty-five months).
Statistical Analysis
Nonparametric statistical analysis was performed with use of SPSS 10.0 for
Macintosh OS (SPSS, Chicago, Illinois) to determine relationships between
variables. Spearman rank correlation tests were used for quantitative data
analysis, and the Kruskal-Wallis test was used for qualitative data analysis.
Significance was set at p = 0.05 and a trend was reported when the p value
was >0.05 but =0.1. In addition, Bonferroni-Dunn corrections were made
in the case of multiple comparisons to exclude a false-positive statistical
correlation.
The intraobserver and interobserver reliabilities of the radiographic and
computed tomography grading systems were calculated with use of variance
component analysis with SPSS 10.0 on a Macintosh OS. The variance component
method was a restricted maximum likelihood estimation.
The radiographic acromiohumeral distance; glenoid retroversion measured on
computed tomography scans; humeral subluxation; radial mismatch of the
components (in millimeters); intraoperative presence of rotator cuff tears;
Goutallier stages of the supraspinatus, infraspinatus, subscapularis, and
teres minor muscles as assessed on postoperative computed tomography scans;
and glenoid inclination were analyzed in relation to the absolute Constant
scores, subjective shoulder values, pain scores, abduction strength scores,
and measurements of active range of motion.
Clinical Results
The mean absolute and relative Constant scores were 39 points
(range, 8 to 77 points) and 50% (range, 9% to 94%), respectively,
preoperatively, and they improved to a mean of 70 points (range, 27 to 95
points) and 85% (range, 30% to 100%), respectively, postoperatively (p =
0.0001) (Table I). The
subjective shoulder value averaged 33% (range, 0% to 85%) preoperatively and
77% (range, 20% to 100%) postoperatively (p = 0.07). The mean pain score
improved from 5 points (range, 0 to 10 points) preoperatively to 13 points
(range, 0 to 15 points) after the operation (p = 0.001). Postoperatively, the
mean forward flexion improved by 34° (p = 0.001) and the mean abduction,
by 46° (p = 0.006). External rotation averaged 16° preoperatively and
45° postoperatively (p = 0.07). With the numbers available, there were no
significant differences in the absolute Constant scores according to the
diagnosis (p = 0.62).
Complications
The complication rate was 19% (nine shoulders), and six of the nine
patients underwent additional procedures. One patient sustained an
intraoperative laceration involving approximately one-third of the diameter of
the axillary nerve, which was repaired intraoperatively. Postoperatively, the
patient had a partial axillary nerve palsy with weakness but activation of the
entire deltoid muscle and eventually recovered 4+ motor function. One patient
had an intraoperative fracture of the greater tuberosity, which was repaired.
There was one postoperative traumatic rupture of the subscapularis with
subsequent repair, and another patient had atraumatic subscapularis
insufficiency that was treated nonoperatively. One patient sustained a
posteroinferior glenoid fracture and underwent intraoperative repair with a
humeral head autograft. One patient sustained an asymptomatic, atraumatic tear
of the rotator cuff, with a decreased acromiohumeral distance seen on the
postoperative radiographs. Another patient sustained a late periprosthetic
fracture of the humeral shaft distal to the lip of the prosthetic stem, which
was treated nonoperatively and healed uneventfully.
Loosening
Two patients had symptomatic loosening of the glenoid component. Both
underwent revision surgery for removal of the glenoid component and autogenous
bone-grafting. Both had a stable humeral component and little evidence of wear
at the articular surface of the glenoid implant. Neither had evidence of
infection. In both patients, the cement was primarily intact around the
glenoid pegs, with the loosening occurring at the bone-cement interface
(Fig. 3). Both patients had a
large humeral head component matched with a medium glenoid component, which
corresponds to a 0-mm radial mismatch. The radiographic lucency scores were 3
and 4, and the computed tomography lucency scores were 7 and 11.
There was a trend for higher computed tomography lucency scores (but not
for higher radiographic lucency scores) to be associated with abnormal
measurements of glenoid version on preoperative computed tomography scans (p =
0.078). With the numbers available, there was no significant association
between component radial mismatch and postoperative computed tomography or
radiographic lucency scores or between the degree of fatty infiltration of the
rotator cuff muscles and lucency scores.
Of the seven patients who had a postoperative computed tomography lucency
score of >6, four had a diagnosis of inflammatory arthritis, two had
primary osteoarthritis, and one had instability arthritis. There was no
significant association between the diagnosis and the lucency scores or
between the diagnosis and the absolute Constant score (p = 0.61).
Radiographic Scoring
Glenoid radiolucencies were observed on the immediate postoperative
radiographs of five shoulders. Twenty-eight shoulders (60%) had grade-A
seating on immediate postoperative radiographs; seventeen (36%), grade-B; one
(3%), grade-C; and one (3%), grade-D. At the time of the most recent
follow-up, twenty-one (45%) of the forty-seven shoulders had radiolucencies
and three (6%) had a Lazarus score of =3
(Table II). Twenty-one of the
forty-seven shoulders were seen to have had progression of the lucency when
the most recent radiographs were compared with the immediate postoperative
radiographs, and progression of at least two grades was seen in nine
shoulders. Two glenoid components were considered to be possibly loose and one
was seen to be definitely loose on radiographs
(Fig. 4). Lucencies were most
often seen initially around the inferior pegs, with higher radiographic scores
for patients with involvement of both superior and inferior pegs.
Computed Tomography Scoring
Computed tomography showed radiolucencies in thirty-six of the forty-seven
shoulders at the time of follow-up. Six shoulders (13%) had a score of >6
points, indicating possible or definite loosening. All six had two or more
zones with at least 1 to 2 mm of radiolucency (Figs.
5 and
6). The areas that were most
often affected were the anteroinferior and posteroinferior pegs (zones 5 and
6) (Table III). In ten
shoulders, computed tomography scans showed the posteroinferior peg (zone 6)
perforating through the posterior glenoid wall after implantation. However,
this had no significant effect on the radiographic Lazarus or computed
tomography lucency scores. With the numbers available, the presence of
radiographic lucency was not significantly correlated with the Constant
scores, pain scores, or active shoulder mobility. The computed tomography
lucency scores were significantly correlated with the radiographic
radiolucency scores (r = 0.58, p < 0.001) and weakly but significantly
correlated with the pain scores (r = -0.31, p = 0.04) and abduction strength
(r = -0.36, p = 0.02).
Because of signal artifact from the humeral component, it was sometimes
difficult to evaluate the superior half of the glenoid component. Optimal
preset contrast settings enhanced visualization in sixteen shoulders (34%),
but the superior baseplate (zone 1) could not be analyzed adequately. In
fourteen shoulders (29%), the superior peg (zone 3) could not be assessed for
lucencies. All other zones were easily visualized. If a zone could not be
adequately assessed, a score was not entered, in order to prevent influence on
statistical analyses performed for each region separately.
Correction of Glenoid Version
Twenty-three shoulders had abnormal measurements of glenoid version (range,
+8° to -35°) on preoperative computed tomography scans. Fifteen of
these measurements were restored to normal (0° to -10°) on the
postoperative computed tomography scans. The remaining eight were restored to
a predetermined version that depended on soft-tissue tension and the severity
of the preoperative version. There was a significant correlation between
higher degrees of preoperative glenoid retroversion and lower subjective
shoulder values (r = -0.39, p < 0.018). There was a trend for abnormal
preoperative glenoid version to be associated with increased computed
tomography radiolucency scores (p = 0.08). With the numbers available, there
was no significant correlation between absolute Constant scores and
radiographic lucency scores. Higher values for glenoid retroversion on
postoperative computed tomography scans were associated with lower
postoperative absolute Constant scores (r = -0.31, p < 0.04).
Subluxation
Five shoulders had preoperative evidence of posterior subluxation of the
humeral head on computed tomography scans. All underwent realignment to
correct glenoid version, and computed tomography showed that four of the five
glenoids were restored to 0° to -10° of retroversion without
subluxation. The remaining glenoid was treated with a planned correction that
reduced the retroversion from -35° to -24°. The patient demonstrated
recurrent static posterior subluxation of the humerus on recent postoperative
computed tomography scans. With the numbers available, there were no
significant correlations between humeral subluxation and Constant scores,
mobility, or radiolucency scores.
Quality of the Rotator Cuff Muscles
Fatty degeneration of grade 2 or higher, as assessed with the technique of
Goutallier et
al.22, was seen in
at least one of the rotator cuff muscles of twenty-one shoulders (45%). There
was no significant correlation between fatty degeneration and overall mobility
or pain. However, postoperative fatty infiltration of the infraspinatus muscle
was significantly associated with the postoperative absolute Constant score
and strength score (p = 0.03). Postoperative fatty degeneration of both the
supraspinatus and the infraspinatus was significantly associated with the
abduction strength score (p = 0.02 and 0.04, respectively). Glenoid version,
the inclination angle, and radial mismatch between the components were not
associated with fatty degeneration scores, with the numbers studied.
Intraobserver and Interobserver Reliability
The reliabilities of the radiographic and computed tomography scoring
systems are summarized in Table
IV. The intraobserver reliabilities of both systems were higher
than the interobserver reliabilities. The overall scoring reliability was
higher for the computed tomography lucency scores than for the radiographic
scores.
Experimental studies have suggested that the holding characteristics
of pegged glenoid components should be superior to those of keeled glenoid
components23-26.
Few investigators, however, have analyzed the in vivo performance of cemented
pegged glenoid components. Romeo et al. found progression of lucency adjacent
to thirteen of nineteen pegged
implants11. Trail
and Nuttall found radiolucencies adjacent to eight of twenty-two pegged
glenoids, a rate that was higher than the rate of radiolucencies adjacent to
the keeled components in their
study13. Our study
demonstrated a relatively high prevalence of radiolucent lines (around
twenty-one of forty-seven components). Most lines had a Lazarus score of
<2, but nine of the forty-seven shoulders had progression of the
radiolucency by at least two grades during the follow-up period. Computed
tomography demonstrated an even higher prevalence of radiolucent lines, which
were seen in thirty-six of the forty-seven shoulders. These results were
observed despite the extensive experience of the surgeons, accurate reaming
for component seating, and use of preparation instruments for the implantation
of pegged glenoid components, which have all been shown to improve glenoid
fixation7,24,25.
There may be several reasons for these results. The use of fluoroscopically
guided radiographs has certainly led to the detection of more radiolucent
lines than was seen in most other studies, purely as a result of the
methodology
employed7,12,27-29.
The use of computed tomography is even more sensitive for detecting
radiolucent lines, and use of this method contributed to the detection of more
radiolucent lines than are usually identified in clinical practice. Throughout
the study, we paid close attention to surgical details such as perfect
preparation of the glenoid
bone24,
pressurization of the
cement15, and use
of implants with an optimal peg design and cement mantle
thickness14.
Conversely, we did not use implants with an optimal radial mismatch. Walch et
al. documented that greater glenohumeral radial mismatch is associated with
fewer radiolucent
lines30. The
average mismatch in our patients was less than the recommended 6 to 10
mm30, and both of
our patients who underwent surgical revision had a radial mismatch of 0 mm.
This series, however, was too small to provide statistical confirmation of the
importance of radial mismatch for the development of radiolucent lines.
Nonetheless, the glenoid components that we currently use were adapted for
greater radial mismatch and may show less loosening than those employed in the
present investigation. Another detail that should be studied further is the
cement that we used. We employed low-viscosity cement so that we could
pressurize it easily, but low-viscosity cement has since proven to perform
less well in vivo than high-viscosity cement in total hip replacement. Thus,
the selection of the type of cement may require further
attention31-33.
It is unknown whether restoration of anatomic glenoid alignment improves
clinical or radiographic
outcomes34,35.
We noted a trend for abnormal preoperative glenoid version to be
associated with higher computed tomography radiolucency scores and poorer
subjective shoulder values. This suggests that a large degree of posterior
glenoid retroversion is a negative prognostic factor even when it is
corrected. Static posterior subluxation was corrected in five shoulders with
reaming, but it recurred in a sixth shoulder in which severe posterior version
had not been completely corrected. Because too few shoulders remained
statically subluxated posteriorly, we were unable to document an effect of
postoperative static humeral subluxation on radiolucency scores.
The relationship between radiographic findings and clinical outcome is of
prime interest. Some authors have reported an association between radiographic
signs of loosening and
pain5, and it is
possible that others could not document such an association because
radiographic assessment is error-prone and underestimates the frequency of
radiolucencies13,27.
In our study, the prevalence of lucencies on computed tomography was even
higher than that on fluoroscopically guided conventional radiographs. The
interobserver and intraobserver reliabilities of computed tomography
assessment were better than those of conventional radiographic scoring. In
this study, pain was significantly associated with computed tomography
radiolucency scores, whereas its association with conventional radiographic
scores did not reach significance. This finding suggests that greater imaging
sensitivity may assist in clarifying this issue. With the advent of new
scatter-reduction computer programs, imaging may be further improved and
computed tomography may become the standard for assessing loosening of glenoid
components. The fact that images were viewed directly on the screen rather
than on hard copies is unlikely to have influenced the
results36.
This study, in which optimal implantation and imaging techniques were used,
showed a high prevalence of radiolucencies around pegged glenoid components.
Computed tomography detects glenoid radiolucencies with better sensitivity and
reproducibility than does conventional radiography and may therefore be a
better imaging modality. When assessed with highly sensitive methods, pegged
glenoids also exhibited progressive radiolucencies. Thus, further efforts to
improve the longevity of glenoid components appear to be warranted. ?