Patients
Thirty-seven consecutive shoulder hemiarthroplasties performed
for the treatment of glenohumeral arthritis with severe cuff deficiency
in thirty-four patients between 1985 and 1996 were identified with
the use of the joint registry at our institution. The senior author (R.H.C.)
had performed thirty-one of the procedures, and three other surgeons
had carried out the remaining six. Patients who were treated with
shoulder hemiarthroplasty in the presence of a rotator cuff tear
but had another underlying diagnosis were excluded from the study.
To be included, a patient had to have had a complete preoperative
evaluation, available operative records, and a minimum of two years
of clinical follow-up. Of the initial group of thirty-four patients,
one died and three were lost to follow-up less than two years after the
operation. The remaining thirty patients (thirty-three shoulders)
were followed for an average of five years (range, two to eleven
years) and represent the study population. The most recent follow-up
evaluation was carried out with a physical examination for twenty shoulders,
a written questionnaire for eight, and a telephone questionnaire
for five.
The average age of the patients at the time of the index arthroplasty
was sixty-nine years (range, fifty to eighty-seven years), and the
average height was 160 cm (range, 134 to 180 cm). There were thirteen
men and seventeen women; two men and one woman underwent bilateral shoulder
arthroplasty. Twenty right shoulders and thirteen left shoulders
were treated, and twenty-four of the shoulders were on the dominant
side. Eleven of the shoulders had undergone a previous procedure.
Four of these shoulders had undergone one previous procedure, six
had undergone two, and one had undergone four. The previous procedures
included rotator cuff repair (performed once in three shoulders
and twice in four); excision of a synovial cyst (performed twice
in the same shoulder); open débridement because of postoperative infection
(performed three times in the same shoulder); and resection of the
distal part of the clavicle, repair of the long head of the biceps,
repair of an acromioclavicular disruption, and open capsular release
(performed once each). Eight of the eleven shoulders with a history of
surgery had undergone an acromioplasty and resection of the coracoacromial
ligament.
Surgical Technique and Operative Findings
All of the operations were performed through a deltopectoral
approach. The subdeltoid bursa was incised vertically from the inferior
edge of the coracoacromial ligament (when present) or the lateral
aspect of the tip of the coracoid. The incision was continued distally,
lateral to the conjoined tendons. The fibrous tissue that was present
at the usual location of the rotator interval and the proximal portion
of the subscapularis tendon were left intact. When passive external
rotation was >30°, the inferior fibrous tissue and any
remaining subscapularis were incised about 1 cm medial to the lesser tuberosity;
otherwise, it was released from the humerus. The anteroinferior
aspect of the shoulder capsule was released from the humerus, and
the humeral head was dislocated anteroinferiorly.
The humeral canal was entered, and the humeral diaphysis was
prepared. With use of a humeral head resection guide with both extramedullary
and intramedullary referencing, the humeral head was resected in
35° of retrotorsion at the level of the previous insertion site
of the supraspinatus tendon. The joint and the rotator cuff could
then be inspected. A massive tear of the rotator cuff was found
in all shoulders. The supraspinatus was torn and the tendon was
absent in every shoulder. The status of the infraspinatus and the
teres minor was graded by assigning the remaining posterior rotator
cuff tissue an hourly position, with twelve o’clock indicating that
it was intact and seven o’clock (for the right shoulder)
or five o’clock (for the left shoulder) indicating that it
was completely absent. This allowed an estimation of the percentage
of the external rotators that was still present: 60% was
present in two shoulders; 40%, in eight; 20%,
in two; and 0%, in twenty-one. The subscapularis was completely
present in sixteen shoulders, it was absent in six, and only the
distal one-half remained in eleven. The biceps tendon was present
but frayed in only seven of the thirty-three shoulders. As noted
before, eight shoulders had undergone a previous subacromial decompression.
Other findings included a fractured acromion, an os acromiale, and
a fracture of the coracoid process.
The humerus was reconstructed with a Cofield implant (Smith and
Nephew Richards, Memphis, Tennessee) in twenty-seven shoulders,
a Bio-Modular implant (Biomet, Warsaw, Indiana) in four, a Neer-II
implant (Kirschner Medical, Fairlawn, New Jersey) in one, and a
Monospherical implant (Howmedica, Rutherford, New Jersey) in one.
Polymethylmethacrylate bone cement was used for stem fixation in
six shoulders. The size of the head was small (<42 mm in
diameter) in three shoulders, medium (42 to 46 mm in diameter) in
twenty-six, and large (>46 mm in diameter) in four. The
median diameter of the humeral head was 44 mm, and the
range was 41 to 54 mm. Glenoid bone loss was described as moderate
(up to the base of the coracoid process) in three shoulders and
severe (medial to the base of the coracoid process) in five. The
glenoid was contoured with a rongeur or burr in six shoulders, and
mechanical reaming of the glenoid and acromion was performed in three;
this allowed better congruency among the osseous glenoid, the undersurface
of the acromion process, and the prosthetic humeral head. Partial
repair of the subscapularis was possible in twenty-three shoulders,
and partial repair of the infraspinatus was possible in three. The
supraspinatus remnants were sutured to the subscapularis muscle
in three shoulders. Eight shoulders had additional procedures, which
included synovectomy (four shoulders), excision of associated synovial
cysts (three shoulders), and a limited acromioplasty (one shoulder).
Postoperatively, the limb was placed in a shoulder immobilizer,
which was used during the daytime and at night for one month. Passive
range-of-motion exercises were started on the day after surgery,
with elevation limited to 120° and external rotation limited to
20°. Pulley exercises for flexion, self-assisted wand
exercises, and isometric strengthening exercises were started at
five to six weeks. Strengthening exercises with an elastic strap were
added at eight to ten weeks after surgery.
Clinical Evaluation
At our institution, the results of clinical assessment of all patients
who have undergone shoulder surgery are recorded on a standard shoulder
analysis sheet that includes entries for pain, range of motion,
and strength10,11. Pain is graded as none, slight, moderate after
unusual activity, moderate, or severe. Active elevation and external
rotation are recorded in degrees, and internal rotation is recorded
as the vertebral segment that the patient can reach with his or
her thumb. Paralysis of the shoulder is given 0 points; a trace
of strength, 1 point; poor strength with gravity eliminated, 2 points; fair
strength against gravity, 3 points; good strength against
resistance, 4 points; and normal strength, 5 points. The overall
results were graded according to the limited-goals criteria proposed
by Neer et al.11 for evaluation
of patients with a deficient rotator cuff. The result was considered
to be successful when the patient had no or slight pain or had moderate
pain only with vigorous activity, had external rotation of >20°, and
had active abduction of >70°.
Radiographic Analysis
The shoulders were evaluated on an axillary radiograph, a 40-degree
posterior oblique radiograph with external rotation of the humerus,
and a 40-degree posterior oblique radiograph with internal rotation
of the humerus. Two shoulders had a complete set of preoperative
and immediate postoperative radiographs but no recent radiographs.
The remaining thirty-one shoulders (94%) had a complete
set of preoperative, immediate postoperative, and most recent radiographs;
the mean duration of radiographic follow-up of these patients was 3.4
years (range, one to ten years).
The preoperative radiographs were used to confirm the diagnosis
of glenohumeral arthritis and to evaluate the degree of bone loss.
They were also assessed to determine if the humeral head was absent
or had collapsed. The acromion was assessed for the presence of
erosion, a fracture, or an os acromiale. Glenoid bone
loss was classified as central or superior and as mild, moderate, or
severe. The distance between the acromion and the humeral head was
measured in millimeters on the external rotation radiograph. Finally,
glenohumeral subluxation was evaluated with regard to the direction
and according to the amount of translation of the humeral head (mild,
moderate, or severe).
The postoperative radiographs were used to evaluate progression
of bone loss and subluxation of the prosthesis, with use of the
criteria described above, and to assess the fixation of the component.
The humeral interface was divided into eight zones12. The presence of radiolucent lines
was recorded, and their zonal location as well as their maximum
thickness within 0.5 mm were noted. Zonal sclerosis and a shift
in component position were recorded as present or absent.
Statistical Methods
Preoperative versus postoperative changes in continuous or ordinal
variables were assessed with the Wilcoxon signed-rank test. Comparisons
of pairs of categorical variables were performed with the Fisher exact
test. Comparisons of a categorical variable paired with either an
ordinal or a continuous variable were performed with the Wilcoxon
rank-sum test. Comparisons of pairs of ordinal or continuous variables
were performed with the Spearman rank correlation coefficient. Logistic
regression was used to assess whether head size was associated
with various outcome measures (mild or moderate instability, greater
than mild pain, migration, and an unsuccessful Neer rating) adjusted
for the height of the patient. Data are presented as the mean and
standard deviation unless otherwise specified. A significance
level of 0.05 was used for all comparisons.
Complications and Reoperations
One patient had an intraoperative nondisplaced fracture of the
medial aspect of the humeral diaphysis that healed uneventfully.
Another shoulder required surgical evacuation of a hematoma twenty-four
days after the index arthroplasty. One of the patients who had undergone
two attempted rotator cuff repairs prior to the index hemiarthroplasty
had increased pain and decreased shoulder strength after an initially
satisfactory postoperative recovery. He was treated with another
partial rotator cuff repair four months after the index arthroplasty.
Three years later, he had moderate pain after unusual activities,
45° of active elevation, 10° of active external rotation, and an
unsuccessful result. There were no other complications, and none
of the patients required revision of the component or another reoperation.
Clinical Results
Shoulder hemiarthroplasty was significantly associated with pain
relief. The mean score for pain decreased from 4.2 points before
the operation to 2.2 points at the most recent evaluation (p = 0.0001).
Preoperatively, pain was severe in eleven shoulders, moderate in nineteen,
and occurred mainly after activity in three. At the most recent
follow-up evaluation, six shoulders had no pain, eighteen had slight
pain, and nine (27%) had moderate pain at rest (four) or
pain with activity (five). Of the four shoulders with moderate pain,
one had progressive proximal migration and an acromial fracture,
two had anterosuperior instability on physical examination, and one
was not followed radiographically and an explanation for the pain
could not be identified.
Active elevation increased significantly in most patients. The
mean active elevation improved from 72° (range, 30° to 150°) before
the hemiarthroplasty to 91° (range, 40° to 165°) at the most recent
follow-up evaluation (p = 0.008). At the most recent evaluation,
three shoulders had 150° of active elevation; three, 120° to 149°; twelve,
90° to 119°; nine, 70° to 89°; and six, <70°. The mean
internal rotation improved from the third lumbar level (range, the
iliac crest to the sixth thoracic level) preoperatively to the first
lumbar level (range, the sacrum to the fourth thoracic level) at
the most recent follow-up evaluation (p = 0.02).
The mean active external rotation improved from 36° (range, 0° to
80°) to 41° (range, 0° to 90°); however, with the numbers available, this
change was not found to be significant (p = 0.1).
Strength in external rotation increased significantly from a
median of 3 points (range, 2 to 5 points) preoperatively to a median
of 4 points (range, 3 to 5 points) at the most recent follow-up
evaluation (p < 0.01). No difference was detected between
the strength in abduction or flexion before the shoulder
arthroplasty and that after it. The median strength in abduction
was 4 points (range, 2 to 5 points) preoperatively and 4 points (range,
3 to 5 points) at the most recent evaluation. The median strength
in flexion was 4 points (range, 3 to 5 points) preoperatively and
4 points (range, 3 to 5 points) at the most recent evaluation.
The overall result was graded as successful in twenty-two shoulders
(67%) and unsuccessful in eleven. The reasons for an unsuccessful
result were moderate pain in three shoulders; moderate pain, active
elevation of <70°, and active external rotation of <20°
in one shoulder; active elevation of <70° in two shoulders;
active external rotation of <20° in two shoulders; and
active elevation of <70° and active external rotation of <20°
in three shoulders. The eight shoulders that had an unsuccessful
result because of poor motion had had a mean of 47° (range, 30°
to 75°) of elevation and 24° (range, 0° to 60°) of external rotation
preoperatively. According to the patients’ assessments,
fifteen shoulders were much better than before the operation, fourteen
shoulders were better, and four shoulders were the same or worse because
of persistent moderate pain.
Of the twenty shoulders that were physically examined at the
time of the most recent follow-up, seven had mild-to-moderate anterosuperior
instability. Four of them had undergone prior surgery involving
the rotator cuff once (two shoulders) or twice (two shoulders).
Two of the seven shoulders had no pain, three had mild pain, and the
other two had moderate pain. Active elevation was £90°
in all but one of them. The Neer rating was successful for five
of these shoulders and unsuccessful for two.
Radiographic Analysis
Preoperatively, the humeral head had collapsed in eight shoulders.
Twenty-five shoulders had superior glenoid bone loss, which was
graded as mild in nineteen, moderate in two, and severe in four.
All shoulders had superior subluxation, which was graded as mild
in two, moderate in six, and severe in twenty-five. In addition,
eleven shoulders had mild and three had moderate anterior subluxation.
The acromiohumeral distance was 0 mm (no separation
between the humeral head and the acromion) in twenty-three shoulders,
2 mm in four, 3 mm in four, 4 mm in one, and 5 mm in one. The acromion
had mild erosion in twenty-one shoulders, had moderate erosion in
five, and was fractured in one. One shoulder had an os acromiale.
Immediately postoperatively, superior subluxation was mild in
one shoulder, moderate in nine, and severe in twenty-three. Fourteen
shoulders had mild anterior subluxation, and three had moderate
anterior subluxation. At the most recent radiographic evaluation,
nineteen shoulders had the same amount and direction of subluxation
as they had had immediately postoperatively (Figs. 1-A, 1-B, and 1-C). Superior subluxation
increased from moderate to severe in seven shoulders between the
immediate postoperative and most recent examinations.
Anterior subluxation increased in one shoulder with and in two without
increased superior migration. The anterior subluxation increased
from mild to moderate in two of these shoulders and from none to
mild in one shoulder. Two shoulders did not have recent radiographs.
At the most recent radiographic evaluation, eight shoulders had
progressive superior erosion of the glenoid (Figs. 2-A, 2-B, and 2-C), fourteen had
progressive erosion of the acromion (Figs. 2-A, 2-B, and 2-C), and two had an acromial fracture.
In addition, eight shoulders had notching of the medial aspect of
the proximal part of the humerus at the level of the inferior
rim of the glenoid (Figs. 3-A, 3-B, and 3-C). Nineteen had no changes at the
periprosthetic interface, three had a 1-mm radiolucent line in one
zone (one shoulder) or two zones (two shoulders), and one had a 2-mm
radiolucent line in two zones. Ten shoulders had reactive sclerotic
lines in one zone (eight shoulders) or two zones (two shoulders).
None of the humeral components was considered to be loose radiographically.
Associations
A history of subacromial decompression was significantly associated
with clinically detectable instability (p = 0.04)
and less active elevation at the time of follow-up (79° ±
37° compared with 95° ± 31° for the shoulders without
a history of subacromial decompression; p = 0.03). Clinically
detectable instability was also significantly associated with less
damage of the infraspinatus and teres minor at the time of surgery
(p = 0.005) and with a surgical attempt to partially repair
the rotator cuff (p = 0.04). A decreased distance between
the humeral head and the acromion was a significant predictor of
decreased strength in external rotation at the most recent follow-up evaluation
(p = 0.03). Progression of bone loss at the scapula or
the proximal part of the humerus was not significantly associated with
longer radiographic follow-up, pain, or an unsuccessful Neer rating.
No significant associations were found between the size of the humeral
head and pain, range of motion, instability, or Neer rating. Although
the size of the humeral head could have been expected to be related
to the size of the patient to some extent, no association was found
between patient height and humeral head size. In addition, humeral
head size was not found to be associated with any of the outcome
parameters when corrected for the height of the patient.
Glenohumeral arthritis in the presence of severe rotator cuff
deficiency is one of the most difficult conditions to treat with
a shoulder arthroplasty4,13. In
the article in which cuff-tear arthropathy was first formally described,
Neer et al. stated that total shoulder arthroplasty had helped all
but one of twenty-six shoulders1.
However, some years later, Franklin et al.14 and
Hawkins et al.15 noted increased
glenoid loosening rates when total shoulder arthroplasty had been
performed in the presence of a deficient rotator cuff. Brownlee
and Cofield13 and Lohr et al.4 compared the results of total shoulder
arthroplasty with those of hemiarthroplasty in patients with cuff-tear arthropathy
and concluded that the former procedure provided better pain relief
but was associated with a substantial rate of glenoid failure. In
contrast, Pollock et al.7 reported
that hemiarthroplasty provided similar pain relief and better active
elevation than did total shoulder arthroplasty in a series
of seventeen shoulders with inflammatory arthritis and thirteen
with cuff-tear arthropathy. Although bipolar hemiarthroplasty has
been used for the treatment of cuff-tear arthropathy16, the results have not been reproducible17.
For the reasons stated above, at the present time many consider
hemiarthroplasty to be the procedure of choice for patients with
glenohumeral arthritis, severe cuff deficiency, and adequate deltoid
function2,5-9. However, shoulder
hemiarthroplasty has not provided consistent pain relief in patients
with this condition4-6,9,10. In
addition, previous authors have not analyzed in detail the relationship
between the outcome of the operation and the status and management
of the coracoacromial arch, the size of the prosthetic head, the
extent and reconstruction of the cuff tissue, and the potential
for postoperative instability and progressive migration and bone
loss. Our study confirmed that shoulder hemiarthroplasty provides
satisfactory pain relief in about three-quarters of patients and
moderate gains in range of motion and strength. It also showed that
clinically detectable instability or progression of scapular and humeral
bone loss are to be expected in a substantial number of patients.
Finally, it identified two factors that were associated with a less
satisfactory outcome: prior subacromial decompression and the extent
of proximal migration of the humeral head.
The reported results of shoulder hemiarthroplasty in patients
with glenohumeral arthritis and a deficient cuff have not been uniform
(Table I). The
percentage of patients with no or mild pain at the most
recent follow-up evaluation after hemiarthroplasty has ranged from
47% to 86%5,6,8,9,
although the rate of subjective satisfaction generally has been
higher. Shoulder hemiarthroplasty has also provided moderate gains
in motion and strength5-9. On
the basis of the limited-goals criteria described by Neer et al.11, a successful result was achieved
in 86% of the shoulders reported on by Williams and Rockwood8 and 63% of the shoulders
in the study by Field et al.6. In our series, 73% of the
shoulders had no or mild postoperative pain and 67% were
rated as having a successful result.
Shoulder instability was reported in all but one of the previously
mentioned studies. In the series of Arntz et al.5,
one shoulder had a reoperation because of symptomatic anterior instability.
Four of the sixteen shoulders reported on by Field et al.6 had postoperative instability, and
two of them required additional surgery. Instability was also noted
in one of the shoulders included in the report by Zuckerman et al.9 and in seven of the twenty shoulders
that were examined at the time of the most recent follow-up in our
study. In addition, we identified progressive migration of the humeral
head in nine of the thirty-one shoulders with complete radiographic
follow-up. As noted by Wiley18 and
by Field et al.6, prior surgical
violation of the anterior aspect of the acromion and the coracoacromial
ligament seems to be the main predisposing factor for this complication.
In our study, a history of subacromial decompression was significantly
associated with clinically detectable instability (p = 0.04)
and less active elevation at the time of follow-up (p = 0.03).
Progressive bone loss is another potential problem after hemiarthroplasty
in shoulders with a deficient cuff. Unfortunately, detailed radiographic
information was not included in some of the previously
published studies. Symptomatic glenoid erosion developed
in two of the eighteen shoulders reported on by Arntz et al.5, and a revision was performed to
modify the size or position of the humeral implant; a third shoulder
had radiographic evidence of medial erosion. In our study, progressive
bone loss was seen at the glenoid in eight shoulders and at the
acromion in sixteen, with a resultant fracture of the acromion in
two of them. In addition, notching at the medial aspect of the proximal
part of the humerus, a radiographic finding that has not
been previously described to our knowledge, developed in eight shoulders.
With longer follow-up than the average two to five years reported
so far in the literature, progressive bone loss
may become an important mode of failure of hemiarthroplasty in this
population. In contrast, loosening of the humeral component does
not seem to be a problem.
The ideal size of the humeral head and the management of the
remaining cuff tissue are two technical factors that need to be
addressed at the time of surgery. A larger humeral head may result
in a more stable reconstruction by increasing soft-tissue tension
and fully articulating with the glenoid and acromion. However, the
resultant overstuffing of the joint impairs the ability to partially reconstruct
the anterior aspect of the rotator cuff and theoretically may lead
to increased joint reactive forces with resultant pain and more
rapid bone loss. A medium-size head was used in most of the patients
in our study. With the numbers available, we did not find any association
between the head size and the outcome of the operation. Regarding
the management of the remaining cuff tissue, in our series attempts
to partially reconstruct the cuff were significantly associated
with clinically detectable instability (p = 0.04), as was
less damage to the posterior part of the cuff (p = 0.005). Resulting
imbalanced force couples may explain these two associations, and
our data do not seem to justify the use of a small humeral head
to facilitate cuff reconstruction.
In conclusion, shoulder hemiarthroplasty provides marked pain
relief in about three-quarters of patients with glenohumeral
arthritis and severe rotator cuff deficiency. It also provides modest
improvements in range of motion and strength. Persistent pain, anterosuperior instability,
and progressive bone loss may complicate the procedure. A less satisfactory
outcome should be expected in patients with prior violation of the
coracoacromial arch. The use of either a small humeral head in an
attempt to facilitate reconstruction of the cuff or a large head
to maximize joint stability does not seem to be justified. Although
shoulder hemiarthroplasty is not a perfect solution for patients
with glenohumeral arthritis and severe cuff deficiency, it probably
represents the best available reconstructive option for this difficult problem
at the present time.