Humeral head replacement was introduced by Neer for the treatment of
selected proximal humeral
fractures1, and
there is a consensus that, in the presence of intact glenoid cartilage,
humeral head replacement is satisfactory for achieving pain relief, stability,
and varying degrees of
movement2,3.
Additionally, humeral head replacement without placement of a glenoid
component has been used to treat various inflammatory, degenerative, and
posttraumatic conditions of the glenohumeral joint in which the glenoid
cartilage has been destroyed. In 1974, Neer reported on the use of humeral
head replacement for the treatment of osteoarthritis; forty-eight patients had
pain relief, and most had an excellent or satisfactory
outcome4. Orfaly et
al. reported pain relief and improvements in motion and function in a study of
twenty-eight humeral head replacements performed for the treatment of primary
osteoarthritis5, and
Edwards et al. reported improvement in all tested parameters, with
seventy-seven excellent-to-good results overall, following eighty-nine humeral
head replacements6.
In counterpoint, several authors have observed less satisfactory outcomes in
the presence of uneven glenoid
erosion7-9.
Also, in a study comparing total shoulder arthroplasty with humeral head
replacement, Gartsman et al. noted improvement, as in earlier studies, but
reported that three of twenty-four shoulders treated with humeral head
replacement alone underwent revision because of progressive glenoid wear and
pain10.
We have also noted variable outcomes in several patient groups. In a 1986
preliminary report on thirty-nine shoulders treated with humeral head
replacement because of glenohumeral arthritis, eight underwent revision
surgery because of painful glenoid
arthrosis11. This
was followed by a 1995 report on thirty-five shoulders with osteoarthritis
followed for an average of 9.3 years after humeral head
replacement12. Pain
relief was maintained in nineteen shoulders, with seventeen having an
excellent or satisfactory result, but nine shoulders required revision because
of painful glenoid arthrosis.
The purpose of this study was to review our continuing experience with
humeral head replacement for the treatment of primary osteoarthritis. With a
minimum five-year follow-up, this study expands on our previous studies
through the evaluation of a more uniform patient population and with ongoing
follow-up examinations.
Three surgeons at our institution performed a total of sixty humeral head
replacements for the treatment of primary osteoarthritis of the glenohumeral
joint between November 8, 1978, and January 15, 1997. Patients were identified
with the use of a computerized database containing the files of all patients
who had undergone joint arthroplasty at our institution. The study was
approved by our institutional review board. Patients entered in this database
are asked to return for an interview, examination, and radiographic evaluation
at regular follow-up intervals. Patients who are unable to return are sent a
questionnaire and are requested to have a local orthopaedic surgeon send the
results of a clinical examination as well as recent radiographs. Five patients
(seven shoulders) died less than five years postoperatively, and one patient
(two shoulders) was lost to follow-up four years following the initial
arthroplasty. Forty-nine patients with a total of fifty-one humeral head
replacements had a complete postoperative evaluation and were followed for a
minimum of five years or until revision surgery. These patients were included
in the clinical analysis. Thirteen of these patients, operated on between 1978
and 1983, were included in a previous study initiated in
198511 and
published in
199512. At the time
of that report those patients had been followed for an average of ten years
and one month, whereas at the time of the current study they had been followed
for an average of fifteen years. The average duration of clinical follow-up
for all fifty-one shoulders was 11.3 years; it was less than five years for
two shoulders (treated with revision surgery), between five and ten years for
twenty, more than ten but less than fifteen years for sixteen, more than
fifteen but less than twenty years for six, and more than twenty but less than
twenty-five years for seven. Follow-up radiographs were available for
forty-two shoulders, and a complete series of radiographs (made
preoperatively, early postoperatively, and at the time of the most recent
follow-up [at an average of 10.7 years]) was available for thirty-nine
shoulders.
During the time period of the study, 541 shoulders were treated with total
shoulder arthroplasty because of osteoarthritis. A retrospective analysis
suggested that the indications for isolated humeral head replacement were a
relatively younger age or a relatively higher activity level in the presence
of a glenoid with minimal bone erosion, a minor amount of uneven bone erosion
that could be easily corrected to neutral glenoid version, or a glenoid with
central bone erosion that would make secure fixation of a glenoid component
difficult (present in thirteen of the twenty patients who were sixty years of
age or older). The mean age at the time of the humeral head replacement was
fifty-nine years (range, thirty-seven to eighty-three years), with one
shoulder in a patient between thirty and thirty-nine years old, eight in
patients between forty and forty-nine, twenty-two in patients between fifty
and fifty-nine, six in patients between sixty and sixty-nine, thirteen in
patients between seventy and seventy-nine, and one in a patient older than
eighty. There were thirty men and nineteen women; one man and one woman
underwent bilateral humeral head replacement. Only one of the fifty-one
shoulders had undergone a previous procedure—an open acromioplasty and
distal clavicular resection. Four of the operatively treated shoulders were in
homemakers, sixteen were in office workers, four were in patients who
performed light manual labor, and eleven were in patients who performed heavy
manual labor. The occupations had not been recorded for the remaining
patients.
Pain was evaluated on a five-grade
scale13,14.
The pain was not graded as 1 (no pain), 2 (slight pain), or 3 (moderate pain
after unusually vigorous activities) in any shoulder preoperatively. Twenty
shoulders had grade-4 pain (moderate pain at rest or with ordinary activities)
and thirty-one, grade-5 (severe pain at rest or with activities)
preoperatively. Active abduction averaged 104° preoperatively; external
rotation, 27°; and internal rotation, to L4-L5.
During the surgery, a rotator cuff tear was identified in ten shoulders.
All tears were small to medium (<1 to <3 cm in diameter) and confined to
the supraspinatus tendon. All were repaired at the time of the humeral head
replacement. An acromioplasty was performed in four shoulders, and a distal
clavicular excision was performed in four. The posterior aspect of the
shoulder capsule was tightened in one shoulder. The glenoid was usually left
undisturbed; however, in nine shoulders it was reshaped to a slight concavity
with either a burr (five shoulders) or a pediatric acetabular reamer (four
shoulders). The humeral head component was press-fit in forty-seven shoulders,
and supplemented fixation was used in four shoulders (bone-grafting of the
humeral shaft in three and bone cement in one). A smooth-textured Neer-II
component (3M, St. Paul, Minnesota) was used in twenty-nine shoulders; a
Cofield tissue-ingrowth component (Smith and Nephew, Memphis, Tennessee), in
eighteen; and a Bio-Modular light and medium textured stem (Biomet, Warsaw,
Indiana), in four (Figs. 1-A and
1-B, 1-C and
1-D).
A complete series of radiographs was available for thirty-nine shoulders
with a mean follow-up period of 10.7 years. The radiographs used for analysis
included 40° posterior oblique views with internal and external rotation
of the humerus and an axillary view. All shoulders had loss of glenoid
cartilage from the area of glenoid contact with the native or prosthetic
humeral head. Glenoid erosion was graded as none when the original subchondral
plate was clearly visible, as mild when a portion or all of the subchondral
plate was eroded but to a depth of <5 mm, as moderate when the erosion
approached the lateral aspect of the base of the coracoid process (typically 5
to 10 cm of erosion), and as severe when the erosion was beyond the lateral
aspect of the base of the coracoid (typically >10 mm in depth).
Preoperatively, glenoid erosion was graded as none in eight shoulders, mild in
twenty, moderate in four, and severe in seven. A biconcave glenoid cavity was
present preoperatively in nine of the thirty-nine shoulders—all with
mild glenoid erosion and mild (thirty-six) or moderate (three) posterior
humeral subluxation.
Glenohumeral subluxation was evaluated with regard to its direction and
degree and was graded as none, mild (when the center of the humeral head or
prosthetic humeral head was translated <25% relative to the center of the
glenoid), moderate (when the center of the humeral head or prosthetic humeral
head was translated 25% to 50% relative to the center of the glenoid), or
severe (when the center of the humeral head or prosthetic humeral head was
translated >50% relative to the center of the glenoid). Preoperatively,
glenohumeral subluxation was present in twenty-five of the thirty-nine
shoulders, and it was more than mild in twelve of them. Postoperatively,
glenoid osseous erosion and subluxation were evaluated in a similar fashion on
the most recent radiographs available.
Periprosthetic lucency and shifting of the humeral component were evaluated
by three of us (D.M.R., J.W.S., and R.H.C.), and a consensus was reached.
Periprosthetic lucency was graded as none, 1-mm thick and incomplete, 1-mm
thick and complete, 1.5-mm thick and incomplete, 1.5-mm thick and complete, or
2-mm thick and complete. A shift in component position was graded as either
present or not present on the basis of a comparison of the early postoperative
and most recent radiographs.
The clinical results were graded with use of a modification of the Neer
result rating
system10,11.
An excellent result indicated that the patient had no or slight pain, was
satisfied with the procedure, and had active abduction of =140° and
external rotation of =45°. The result was rated as satisfactory result
when the patient had no pain, slight pain, or moderate pain only with vigorous
activities, was satisfied with the procedure, and had active abduction of
=90° and external rotation of =20°. The result was considered to
be unsatisfactory when these criteria were not met or the patient had
undergone an additional operative procedure.
All of the original sixty humeral head implants were included in a survival
analysis. Survival free of revision surgery (a change in or the addition of
arthroplasty components) was estimated as a function of the time since the
index procedure with the use of the Kaplan-Meier method and the Cox
proportional hazards survival method with adjustment for correlated data for
two shoulders in the patients with bilateral humeral head replacement.
Univariate associations of risk factors with revision-free survival were
assessed with log-rank tests. Associations between dichotomized risk factors
and pain or movement were assessed with two-sample t tests. Associations
between discrete risk factors and the Neer rating were assessed with an
extension of the Fisher exact test for ordered contingency tables. A p value
of <0.05 was selected as the level of a significant difference among the
various parameters.
Complications, Revisions, and Survivorship
Five patients had a perioperative complication not requiring additional
treatment. These complications included four transient neuropathies and one
episode of postoperative aspiration pneumonia. Ten (17%) of the sixty humeral
head replacements were revised. One patient had a perioperative posterior
dislocation and underwent revision surgery five months after the humeral head
replacement. The other nine revisions (conversions to total shoulder
arthroplasty) were performed because of painful glenoid arthrosis at a mean of
7.6 years (range, 2.3 to 14.3 years) postoperatively (Figs.
2-A and 2-B,
2-C and 2-D,
2-E). There were no revisions
in the shoulders with a repaired rotator cuff tear, and none of the shoulders
had a full-thickness rotator cuff tear at the time of the revision surgery.
Also, only two shoulders had more than mild subluxation, which was posterior
in one and anterior in the other, at the time of the revision surgery. The
estimated survival rate, with 95% confidence intervals, of the humeral head
replacements was 95% (89% to 100%) at five years, 86% (76% to 96%) at ten
years, 73% (58% to 91%) at fifteen years, and 73% (53% to 91%) at twenty years
(Fig. 3). Age as a continuous
variable was significantly associated with revision (p = 0.004). The hazard
ratio of 0.935 (95% confidence interval, 0.893 to 0.979) indicates that older
age is protective against revision surgery. When age was dichotomized (as an
age of fifty-nine years or younger or an age of sixty years or older), it did
not appear to be a significant variable (p = 0.21). However, in this analysis
the hazard ratio was 0.380 (95% confidence interval, 0.083 to 1.751), again
indicating that older age was protective against revision surgery. With the
numbers studied, there were no significant differences in survivorship free of
revision surgery according to gender, the presence of a rotator cuff tear, the
presence or amount of glenoid erosion, the presence or absence of a biconcave
glenoid, or the presence or amount of joint subluxation.
Clinical Evaluation
There was significant pain relief following the surgery in the fifty-one
shoulders, with the most recent mean pain score decreasing to 2.4 points from
a mean of 4.6 points preoperatively (p < 0.0001). The distribution of pain
ratings preoperatively and at the time of the most recent follow-up (mean,
11.3 years) is shown in Figure
4. Importantly, just prior to revision surgery or at the time of
the most recent postoperative assessment, moderate pain was recognized in nine
shoulders and severe pain, in seven shoulders. With the numbers studied, no
significant associations were detected between age, gender, a rotator cuff
tear, glenoid erosion, biconcavity, or subluxation and either the preoperative
or the postoperative pain score.
At the time of the most recent follow-up (at a mean of 10.7 years) for the
shoulders with complete radiographic evaluation, the mean active abduction was
134° (range, 40° to 180°), the mean external rotation was 49°
(range, —5° to 100°), and the median postoperative internal
rotation was to L2 (range, the trochanter to T5). There was significant
improvement in active abduction (p < 0.0001), external rotation (p <
0.0001), and internal rotation (p < 0.024) following the surgery. With the
numbers studied, abduction, external rotation, and internal rotation were not
significantly associated with age at the time of the index operation, gender,
the presence of a rotator cuff tear, preoperative glenoid erosion,
preoperative glenoid biconcavity, or immediate postoperative subluxation.
Twenty-four of the thirty-nine humeral head replacements for which the most
recent radiographs could be evaluated had some degree of glenohumeral
subluxation, which was more than mild in ten of them. Glenoid erosion was
present in thirty-seven of the thirty-nine shoulders. It was mild in fourteen
shoulders, moderate in eleven, and severe in twelve. Glenoid erosion had
increased between the preoperative and most recent radiographic evaluations (p
< 0.0001). Six shoulders had had correction of glenoid biconcavity during
the humeral head replacement: three went on to revision, and three did not. Of
the unrevised shoulders with correction of the biconcavity, two had an
unsatisfactory result and one, a satisfactory result. The three shoulders that
had biconcavity but no correction did not undergo revision, and all had a
satisfactory result. Glenoid biconcavity developed postoperatively in two
shoulders, both of which underwent revision. Humeral periprosthetic lucency
was present in thirteen of thirty-nine shoulders. It was 1-mm thick and
incomplete in seven shoulders, 1-mm thick and complete in one shoulder, 1.5-mm
thick and incomplete in two shoulders, and 1.5-mm thick and complete in three
shoulders (each with a Neer-II component). No shoulder had a 2-mm-thick
complete line. There was a shift of the position of one Neer-II humeral
component.
Overall, there were ten excellent, twenty satisfactory, and twenty-one
unsatisfactory results. The reason for the unsatisfactory results was revision
surgery in ten shoulders, moderate or severe pain in five, a lack of abduction
(<90°) in three, a lack of external rotation (<20°) in two, and
pain and a lack of active abduction in one.
The findings of our study are in agreement with those of previous studies
showing that humeral head replacement for the treatment of osteoarthritis
usually provides significant long-term pain relief as well as improvement in
active abduction, internal rotation, and external
rotation6,10,15.
Humeral head replacement has been noted to be associated with decreased
overall initial costs, shorter operative times, and decreased blood loss when
compared with total shoulder
arthroplasty10.
However, in this study, the survival of humeral head replacements was inferior
to that of total shoulder replacements, for which the survival rates were
reported to be 97% at five years, 97% at ten years, and 84% at fifteen years
in one study15 and
93% at ten years and 87% at fifteen years in another
study16. Ten of the
sixty shoulders in our series underwent revision surgery, with nine of the ten
revisions being specifically due to pain secondary to glenoid arthrosis.
Nine shoulders were identified as having a biconcave glenoid
preoperatively, but the amount of eccentric glenoid wear was thought to be
mild in these nine shoulders. With only mild eccentric wear being present, it
is understandable how our results would be contrary to those in previous
studies7 in which
the results of humeral head replacement in the presence of a biconcave glenoid
were inferior to the results of humeral head replacement in the presence of a
concentric glenoid. In this study, neither subluxation nor glenoid erosion was
found to be significantly associated with pain relief or an excellent or
satisfactory outcome.
In some prior studies, patients with rotator cuff lesions were excluded
from the evaluated
population5,7,10,
whereas others included repairable
tears6,8,15,17.
A rotator cuff tear was present in ten of the fifty-one shoulders in our
series. The tears were small to medium and confined to the supraspinatus
tendon. The inclusion of patients who had had a repair of a small-to-medium
rotator cuff tear was based on the previous finding that repairable rotator
cuff tears do not significantly affect the results of shoulder
arthroplasty6,9.
We currently believe that humeral head replacement is indicated for
patients with osteoarthritis who have severe central glenoid erosion and those
who are younger and more active and wish to continue high-demand vocational or
avocational pursuits. The latter group must be willing to accept the
possibility that, in exchange for the benefits of being able to maintain a
more active and rigorous lifestyle, they may not obtain complete pain relief.
We perform humeral head replacement only in patients who have a balanced joint
or a joint that can be balanced with a minimal amount of glenoid reaming. In
line with this, our current techniques include limited adjustment of humeral
head rotation: it is varied between 20° and 35° as a function of
preoperative humeral subluxation, and glenoid erosion is corrected with use of
a glenoid reamer that matches the surface shape of the humeral component when
the glenoid is irregular in shape or is flat. Correction of glenoid version is
limited to a few millimeters and certainly to <5 mm. We continue to
consider this procedure for patients with smaller and medium rotator cuff
tears, with rotator cuff repair performed at the time of the
surgery12. Also, we
perform intraoperative capsular releases to obtain normal or nearly normal
passive movement, and postoperatively we make great efforts to avoid stiffness
as we believe that doing so may be associated with a higher probability of
pain relief.
To our knowledge, this is the only long-term study of humeral head
replacement from a single institution. It supports the earlier findings of
significant reduction in shoulder pain and improvement in range of motion, but
there is a relatively high rate of unsatisfactory results, including the need
for revision surgery. Careful consideration should be given to whether this is
the optimal procedure for treatment of osteoarthritis of the shoulder.
?