All patients gave informed consent prior to the procedure, and the study
received institutional review board approval. Twelve surgeons practicing in
twelve different centers participated in the study, which was prospective. The
numbers of arthroplasties performed by each surgeon ranged from one to
sixteen. The selection of patients for hemiarthroplasty or total shoulder
arthroplasty was left to the discretion of the treating surgeon, who employed
established historical guidelines in the selection process
(Table I). Patients with an
irreparable rotator cuff tear or marked attenuation of the rotator cuff were
not treated with a total shoulder arthroplasty.
All patients in the study had inflammatory arthritis of the shoulder. No
patient who had had insertion of a shoulder implant by any of the twelve
surgeons during the time frame of the study was excluded.
A single shoulder implant system (Global; DePuy Orthopaedics, a Johnson and
Johnson Company, Warsaw, Indiana) that incorporated modular humeral components
of various sizes designed for press-fit or cemented fixation was used.
Prosthetic, noneccentric humeral heads were available in multiple sizes, each
to be used with an all-polyethylene glenoid component with a mismatch of 6 mm
between the radius of its articular surface and the radius of the
corresponding head. The glenoid components had either five pegs or a central
keel that enabled implant fixation to bone with cement.
Standardized forms were utilized to record clinical information including
demographic data; the patient's self-assessment of comfort and quality of life
on a visual analogue scale, the Simple Shoulder Test, and a forty-eight-item
activities questionnaire; and the results of a physical examination that
documented range of motion, strength, and stability.
Radiographic Assessment
Radiographic evaluation included anteroposterior and transaxillary lateral
views. Other than the utilization of these two standardized views, no formal
controls for patient positioning were utilized during the radiographic
examination. Clinicians participating in the study submitted radiographs for
review that were deemed to be the highest quality obtainable.
X-ray-compensating filters were not employed for the anteroposterior
radiographs.
All submitted radiographs were analyzed in the same manner. The
preoperative radiographic analysis included assessment of disease severity and
measurement of the acromiohumeral interval and the humeral head offset. The
acromiohumeral interval was measured on the anteroposterior radiograph,
between the inferiormost projection of the acromion process and the
superiormost projection of the anatomic or prosthetic humeral head. Humeral
head offset was measured, also on the anteroposterior radiograph, from the
medial aspect of the base of the coracoid process to a point on the line
defining the center of the medullary canal of the humerus.
In addition to humeral head offset and the acromiohumeral interval, the
postoperative radiographic analysis included measurements of humeral head
subluxation, humeral head height, adaptation of the prosthesis to the anatomy,
humeral component fit, humeral component orientation, and glenohumeral
alignment.
Humeral head subluxation was determined by measuring, on both radiographic
views, the distance between the center of the glenoid and the center of the
humeral head. Subluxation was considered to be present and mild if the
measurement was less than one-half of the humeral head diameter and severe if
it was greater than one-half of the humeral head diameter. Humeral head height
was measured on the anteroposterior radiograph, from the superiormost aspect
of the prosthetic humeral head to the superiormost aspect of the greater
tuberosity. The humeral head was considered to be high if the measurement
exceeded 10 mm, normal if it was <10 mm, and low if it was distal to the
level of the greater tuberosity.
The severity of inflammatory arthritis of the shoulder has been classified
on the basis of radiographic
criteria4,7,8.
This has enabled investigators to stratify treatment protocols and better
understand outcomes. With this in mind, we selected the criteria of
Steinbrocker et al. and further stratified the patients in our
study8. These
criteria were sufficient to define two groups of patients: those with and
those without extremely severe disease. Three criteria had to be present to
classify the disease as extremely severe. First, radiographic evidence of
osteopenia had to be seen. Osteopenia was defined as loss of bone structure
when expressed as a ratio of the measurement of the humeral cortical thickness
to the humeral diameter at the superior aspect of the "humeral
cylinder" (the proximal metaphyseal-diaphyseal junction where the
trumpet-shaped proximal humeral metaphysis ends and the cylindrical diaphysis
begins) on the anteroposterior radiograph. If the cortical
thickness-to-humeral diameter ratio was =0.1, osteopenia was considered to
be present. Second, erosive changes of inflammatory arthritis had to be
present on both sides of the joint with loss of articular architecture for the
disease to be considered extremely severe. Third, gross malalignment
(migration, subluxation, or dislocation) had to be present; malalignment
referred not only to head-glenoid displacement but also to anomalies due to
medialization secondary to central erosion of the glenoid
(Fig. 1).
The term "adaptation to anatomy" is an expression of
discriminating criteria relative to the size and position of the humeral and
glenoid components. Humeral prosthetic adaptation was defined by a humeral
component in neutral alignment with respect to the long axis of the humeral
shaft and a component size that maximally filled the medullary canal on the
anteroposterior and axillary lateral radiographs. The glenoid component was
determined to be adapted to the scapula when the percentage of the component
that was supported by glenoid bone exceeded 75% on both anteroposterior and
axillary lateral radiographs. Since the surgical ideal is to achieve complete
congruency between the component and the host subchondral bone, any section of
a component that was backed by an intervening layer of cement only was deemed
to be unsupported. Prosthetic adaptation to anatomy was said to have occurred
when at least two of three observers determined that both the humeral and the
glenoid component met the above criteria.
Postoperative radiographs were used to determine whether anatomic
glenohumeral alignment had been achieved by the shoulder arthroplasty. As seen
on the anteroposterior radiograph, alignment was considered not to have been
achieved when there was glenohumeral dislocation or subluxation, the head
height was >10 mm or distal to the level of the greater tuberosity, the
acromiohumeral interval was <3 mm, or the glenoid component was positioned
inferiorly. As seen on the axillary radiograph, alignment was considered not
to have been achieved when the glenoid component was tilted anteriorly or
posteriorly >15° from the normal glenoid axis.
Postoperative radiographs were reviewed for six characteristics that were
considered to be evidence of changes over time. These changes were glenoid
erosion to any extent, any decrease in humeral head offset, migration of the
glenoid component, a complete radiolucent line of >1 mm in thickness at the
humeral or glenoid component interface, >50% change in the acromiohumeral
interval, and any change in the head height in relation to the greater
tuberosity reflecting subsidence of the humeral component.
Statistical Methods
T tests were used for comparisons of mean values for interval level data
between the two groups. Unequal variance t tests were used when variances were
significantly different (p < 0.05). Paired t tests were used to determine
differences between preoperative and postoperative values within a group. In
addition to range-of-motion measurements, scores on the Simple Shoulder Test
and the visual analogue scale were considered to be interval level
measures.
Categorical measures were analyzed with use of Fisher exact tests when
there were two groups and two outcomes. Chi-square tests were employed where
there were more than two outcomes. When a categorical measure with more than
two outcomes was tested to determine whether it was associated with
differences in an interval level measure, analysis of variance was used.
Data were gathered prospectively for 144 consecutive shoulders: ninety-two
treated with hemiarthroplasty and fifty-two treated with total shoulder
arthroplasty. Nine patients, eight from the hemiarthroplasty group and one
from the total shoulder arthroplasty group, died and were withdrawn from the
study. Seventy-four shoulders that had been followed for less than twenty-four
months were excluded as well. The study therefore included the remaining
sixty-one shoulders, thirty-six treated with hemiarthroplasty and twenty-five
treated with total shoulder arthroplasty, all of which were followed for at
least twenty-four months. The mean duration of clinical follow-up was
thirty-eight months (range, twenty-four to seventy-three months) after the
hemiarthroplasties and thirty-nine months (range, twenty-four to sixty-nine
months) after the total shoulder arthroplasties. The mean duration of
radiographic follow-up was thirty months (range, twenty-four to seventy-two
months) for both groups.
The underlying diagnoses were rheumatoid arthritis (fifty-three shoulders),
lupus (two), and nonspecific inflammatory arthritis (six). The female-to-male
ratio was three to one. The mean patient age at the time of the arthroplasty
was fifty-eight years (range, thirty to eighty-four years).
Clinical Results
There was significant improvement (p < 0.0001) in all parameters
measured with the visual analogue scale, including overall pain, shoulder
comfort with the arm at rest, shoulder comfort, during sleep, overall shoulder
function, ability to use the arm full-time, and quality of life. With the
numbers available, no significant difference could be detected between the
hemiarthroplasty and total shoulder arthroplasty groups with regard to any of
these parameters.
There was an average improvement of 4.3 "yes" responses on the
Simple Shoulder Test after the operations. This was a significant improvement
overall as well as within the separate total shoulder arthroplasty and
hemiarthroplasty groups (p < 0.0001). With the numbers available, no
significant difference could be detected between the two groups.
Of the forty-eight activities of daily living on the activities
questionnaire, the ten (in descending order) that were determined to be the
most improved according to the patients' self-assessment were the ability to,
without help, put the arm into a coat sleeve, tuck a shirt behind the back,
place the hand behind the head, comb and brush the hair, carry a 10-lb
(4.5-kg) suitcase, get in and out of a low chair, wash the middle of the back,
take clothes from a rack, lift 10 lb to shoulder height, and dress and
undress. The ability to carry out these ten activities was improved for 35% to
46% of the patients after the operation. With the numbers available, no
significant difference could be detected between the patients treated with
total shoulder arthroplasty and those treated with hemiarthroplasty.
The range of motion improved significantly after both the
hemiarthroplasties and the total shoulder arthroplasties (p < 0.05). Active
elevation was significantly better after the total shoulder arthroplasties
than after the hemiarthroplasties (p < 0.004). The hemiarthroplasty group
had a mean of 74° of active elevation preoperatively and 89°
postoperatively, resulting in a mean improvement of 15°. The total
shoulder arthroplasty group had a mean of 69° of active elevation
preoperatively and a mean of 116° postoperatively, with a mean improvement
of 47°. The remaining motion parameters also improved, although no
significant difference could be detected between the two groups, with the
numbers available.
A full-thickness rotator cuff tear was present in 61% (twenty-two) of the
thirty-six shoulders treated with hemiarthroplasty and in 36% (nine) of the
twenty-five treated with total shoulder arthroplasty. The tear was repaired in
all shoulders treated with total shoulder arthroplasty. The preoperative
scores on the visual analogue scale for comfort at rest and during sleep were
significantly higher (more pain) in the presence of a full-thickness rotator
cuff tear (p = 0.009 and p = 0.05, respectively). Postoperatively, the
presence of a repaired or irreparable tear did not influence the scores on the
visual analogue scale or the Simple Shoulder Test, the range of motion, or the
glenohumeral alignment.
Radiographic Results
The disease was extremely severe in sixteen (44%) of the thirty-six
shoulders treated with hemiarthroplasty and nine (36%) of the twenty-five
treated with total shoulder arthroplasty. The presence of extremely severe
disease had no significant effect on the parameters measured on the visual
analogue scale, the responses to the Simple Shoulder Test or the activities
questionnaire, the range of motion, or the alignment of the prosthesis.
Humeral head subluxation was noted after six (17%) of the thirty-six
hemiarthroplasties and after five (20%) of the twenty-five total shoulder
arthroplasties. With the numbers available, the subluxation had no significant
effect on the parameters measured with the visual analogue scale, the
responses to the Simple Shoulder Test or the activities questionnaire, the
range of motion, or the alignment of the prosthesis.
The change from the preoperative to the postoperative humeral head offset
was from 50 to 54 mm in the hemiarthroplasty group and from 52 to 58 mm in the
total shoulder arthroplasty group, an increase of 4 and 6 mm, respectively. A
trend for higher (poorer) pain scores on the visual analogue scale and less
improvement in the Simple Shoulder Test scores (p = 0.067) was observed for
cases with greater humeral head offset. There was a significant association
between increased humeral head offset and decreased passive external rotation
at 90° of abduction in the total shoulder arthroplasty group only (p =
0.02) and between increased humeral head offset and decreased passive
horizontal cross-body adduction in the hemiarthroplasty group only (p =
0.021).
In the hemiarthroplasty group, the head height was >10 mm in eight (22%)
of the thirty-six shoulders, 5 to 10 mm in seventeen (47%), and indeterminate
in eleven (31%). In the total shoulder arthroplasty group, the head height was
>10 mm in eleven (44%) of the twenty-five shoulders, 5 to 10 mm in eleven
(44%), and indeterminate in three. Head height was considered to be
indeterminate when it was not possible to accurately discriminate the
projection of the greater tuberosity from the surrounding soft tissues on the
radiograph. However, the quality of the remaining portion of the radiograph
was satisfactory for analysis of all other radiographic parameters. High
placement of the humeral head in patients treated with a hemiarthroplasty
resulted in significantly poorer mean scores for the ability to use the arm
for work or play on the visual analogue scale (47 compared with 23; p = 0.04)
and on the Simple Shoulder Test (4.5 compared with 7.1; p = 0.016) than when
the head height was normal. No difference was observed for the patients who
had had a total shoulder replacement. In the hemiarthroplasty group, high head
height was also associated with a significant reduction in active total
elevation (76° compared with 100° when the head height was normal; p =
0.036).
The acromiohumeral interval was 3.2 mm preoperatively and 2.7 mm
postoperatively in the hemiarthroplasty group and 3.8 mm preoperatively and
4.0 mm postoperatively in the total shoulder arthroplasty group. Larger
acromiohumeral intervals were associated with superior preoperative scores on
the visual analogue scale for comfort at rest (p = 0.042) and higher
postoperative Simple Shoulder Test scores (p = 0.032), but only in the
hemiarthroplasty group. There was a significant difference in the
postoperative acromiohumeral interval between the shoulders in which
glenohumeral alignment was achieved and those in which it was not achieved
(4.1 compared with 2.2 mm; p = 0.023).
The prosthesis was adapted to the anatomy in nineteen of the thirty-six
shoulders treated with hemiarthroplasty and in twelve of the twenty-five
treated with total shoulder arthroplasty. No significant differences were
noted in the scores on the visual analogue scale or the Simple Shoulder Test
between the shoulders with adaptation of the prosthesis to the anatomy and
those without it. However, whether or not the prosthesis had adapted to the
anatomy had a significant effect on active total elevation (113° compared
with 91°; p = 0.027) and the change in active external rotation (30°
compared with 13°; p = 0.012). Prosthetic adaptation to the anatomy and
restoration of glenohumeral alignment were highly associated with one another
(p < 0.001).
The humeral fit was considered to be undersized after twenty-seven of the
thirty-six hemiarthroplasties and after eight of the twenty-five total
shoulder arthroplasties. In the hemiarthroplasty group, the orientation of the
humeral component was neutral in nineteen shoulders, varus in five, valgus in
three, and indeterminate in nine. In the total shoulder arthroplasty group,
the orientation of the humeral component was neutral in fifteen shoulders,
varus in two, valgus in four, and indeterminate in four. Differences in
humeral component size or orientation did not influence the scores on the
visual analogue scale or the Simple Shoulder Test, the range of motion, or the
glenohumeral alignment.
According to our criteria, glenohumeral alignment was restored in seventeen
of the thirty-six shoulders treated with hemiarthroplasty and thirteen of the
twenty-five treated with total shoulder arthroplasty. Achievement of
glenohumeral alignment resulted in significant improvements and trends toward
improvement in the scores for overall pain (p = 0.10), sleep comfort (p =
0.08), shoulder function (p = 0.08), use of the arm for work and play (p =
0.014), and overall quality of life (p = 0.038) compared with the scores when
alignment had not been achieved. Active total elevation of the shoulders in
which alignment had been restored was significantly greater (by an average of
39°) than that of the shoulders in which alignment had not been restored
(p = 0.0004). Total shoulder arthroplasty significantly outperformed
hemiarthroplasty in shoulders with restoration of glenohumeral alignment, with
a postoperative mean active total elevation of 130° compared with 99°
(p = 0.017) and an improvement from the preoperative status of 59°
compared with 17° (p = 0.022). Passive elevation was also better after
total shoulder arthroplasty than it was after hemiarthroplasty (147°
compared with 115°; p = 0.0063).
No humeral component subsided. Glenoid erosions were noted after four (11%)
of the thirty-six hemiarthroplasties (Figs.
2-A and 2-B), and an insignificant decrease in motion and comfort
was observed to coincide with the onset of the erosion. Two of the glenoid
components loosened after the total shoulder arthroplasties. No other
radiographic changes over time were observed. Overall, the radiographic
changes over time had no significant effect on the scores on the visual
analogue scale or the Simple Shoulder Test, the range of motion, or the
glenohumeral alignment.
With the numbers available, no other meaningful, significant differences or
associations could be derived from the radiographic data, including the effect
of the radiographic parameters on the scores on the activities
questionnaire.
Complications
There were seven complications. Four, including two humeral fractures, were
intraoperative. These nondiaphyseal fractures remained nondisplaced and stable
and did not require additional treatment or alteration of the postoperative
rehabilitation. In one patient, anterior glenohumeral instability was
recognized at the completion of the procedure but before the patient was
awakened, and it was satisfactorily corrected. There was one case of transient
sensory and motor dysfunction of the axillary nerve that resolved without
sequelae within six weeks.
There were three postoperative complications, including a rotator cuff tear
in one patient and signs of impingement in another. In neither instance were
symptoms sufficient to warrant surgical treatment. There was also one regional
pain syndrome that required pain management consultation.
We are not aware of any previous prospective studies in which only patients
with rheumatoid or another form of inflammatory arthritis were treated with
hemiarthroplasty or total shoulder arthroplasty and were compared. The
cumulative data from several studies revealed that, despite a wide spectrum of
disease severity, 521 of 565 patients with inflammatory arthritis of the
shoulder treated with implant arthroplasty had satisfactory, if not excellent,
pain relief at the time of short or medium-term
follow-up9-19.
With longer-term follow-up, pain relief was sustained in 162 of 189
patients4,5,20.
In the study by Boyd et al., there was a trend toward better pain relief after
total shoulder arthroplasty than after hemiarthroplasty, but it was not
significant10. Our
review of the previously published studies identified no relation between pain
relief and functional class, Larsen grade, age, gender, rotator cuff status,
or preoperative or postoperative range of motion. The present study revealed
that favorable clinical and radiographic results can be obtained even for
patients with extremely severe disease.
Contrary to many of the reports in the literature, the majority of the
patients in our study achieved functional recovery, as determined by their own
self-assessment15,17.
Figgie et al. reported that pain relief was good but functional recovery was
inferior for many reasons, including general patient debility, poor soft
tissue and muscle strength, damaged metaphyseal bone stock, and adjacent joint
disease13. Boyd et
al. found a trend toward better functional recovery after total shoulder
arthroplasty than after hemiarthroplasty, but the difference was not
significant10.
Fenlin et al. observed that improved function was less predictably achieved
than was improved
comfort12.
Tears of the rotator cuff are often encountered when patients with
inflammatory arthritis are treated with implant
arthroplasty15,16,21.
There is conflicting information in the literature relative to the presence of
a rotator cuff tear, its reparability, and the effects of each on the outcome
of implant
arthroplasty1,10,13-16.
The present study revealed no differences in the results of total shoulder
arthroplasty performed in patients with an intact rotator cuff compared with
those in patients with a repaired rotator cuff.
Although the criteria for component positioning in the present study differ
somewhat from those of Figgie et
al.13, our data
corroborate theirs. Optimum implant position resulted in less pain and better
function, which perhaps reflected the value of restoring normal anatomy and
mechanics. Furthermore, the selection of appropriately sized implants and the
accuracy of implant placement determine how well the implant adapts to the
anatomy. Appropriate sizing and implantation may influence the clinical and
radiographic outcome of the reconstruction. The present study revealed that
this factor favorably influenced the active range of motion and was associated
with restoration of glenohumeral alignment.
The present study revealed that the differences between the outcomes of
hemiarthroplasty and total shoulder arthroplasty were slight even though the
patients treated with hemiarthroplasty had more severe inflammatory arthritis.
(Extremely severe disease was noted in sixteen of the thirty-six shoulders
treated with hemiarthroplasty as opposed to nine of the twenty-five treated
with total shoulder arthroplasty). Both the absolute recovery and the extent
of recovery of forward elevation were significantly better after the total
shoulder arthroplasties, undoubtedly reflecting a selection bias favoring the
use of total shoulder arthroplasty in the presence of a functional or
reparable rotator cuff. Another factor to consider was the high percentage of
shoulders treated with total shoulder arthroplasty in which anatomic alignment
was restored.
Glenoid erosions were observed after four of the thirty-six
hemiarthroplasties, and this complication remains a concern for patients in
whom inflammatory arthritis is treated with hemiarthroplasty. Although the
outcomes for these patients were not significantly different from those for
the other patients treated with hemiarthroplasty, the onset of erosion
generally heralded the onset of a change in comfort status and a diminished
range of motion. Wakitani et al. observed glenoid erosions after seven of
nineteen
hemiarthroplasties22.
The pattern of destruction did not influence the presence or extent of
erosion. Koorevaar et al. found progressive glenoid erosions in nine of
fifteen patients followed for one to seventeen years (mean, eight years) after
a hemiarthroplasty2.
There was no apparent correlation with the patient's perception of pain.
Although no patient in the present series required a revision procedure
because of intractably painful glenoid erosion, that complication has been
reported to be the most common reason for conversion of a hemiarthroplasty to
a total shoulder
arthroplasty23.
Longitudinal follow-up of this expanding series will hopefully reveal
trends that help to define the role of glenoid resurfacing in patients with
inflammatory arthritis. On the basis of previous experiences and the data
derived from the present study, we recommend glenoid resurfacing when (1) the
rotator cuff is intact or reparable with a reasonable chance for healing and
contribution to function, and (2) there is adequate glenoid bone. The
influence of the patient's age, activity level, general health, and disease
involvement should be considered on an individual basis.
Implant arthroplasty can be contemplated for the treatment of various
stages of inflammatory arthritis of the glenohumeral joint, with the most
favorable outcomes occurring in the presence of an intact rotator cuff and
sufficient bone architecture to support the humeral and glenoid implants.
Improved comfort and functional recovery can be expected for nearly all
patients undergoing hemiarthroplasty or total shoulder arthroplasty. Slightly
better pain scores and forward elevation can be expected after total shoulder
arthroplasty, especially when glenohumeral alignment is restored by the
prosthetic reconstruction. ?
Note: The authors thank the following surgeons for contributing
data for this study: Kenneth W. Bramlett, MD (Birmingham, AL), Jerome V.
Ciullo, MD (Troy, MI), Hal S. Crane, MD (Denver, CO), Donald P. Endrizzi
(Portland, ME), Daniel K. Guy, MD (LaGrange, GA), Robert H. Hawkins, MD
(Vancouver, BC, Canada), Sanford S. Kunkel, MD (Indianapolis, IN), Tom R.
Norris, MD (San Francisco, CA), Charles A. Rockwood Jr., MD (San Antonio, TX),
Anthony A. Romeo, MD (Chicago, IL), and Stephen C. Weber, MD (Sacramento, CA).
The manuscript is dedicated to the contributions and memory of the late
Douglas T. Harryman II, MD.