It is an enormous honor to be asked to give this lecture, particularly when
looking at the roll call of my predecessors. Like most of them, I became
fascinated by the life and work of Sir Robert Jones (1857 to 1933). He was the
nephew of the famous doctor Hugh Owen Thomas. After qualifying as a doctor in
1878, Jones moved to the forefront of hyperspecialism in surgery, as he only
did orthopaedics from 1905 onward. He had a multifaceted career. Because he
had been appalled at the high rate of industrial injury, he developed a
casualty service for workers at the Manchester Ship Canal and saved many
lives. He was the founder of the Robert Jones and Agnes Hunt Hospital,
Director General of Military Orthopaedics at that time, and a founder of the
British Orthopaedic Association, hence this eponymous lecture. He was
obviously extremely forward-thinking and was a pioneer of day surgery. In a
publication in 1908, Robert Jones discussed his eight rules for tendon
transplantation and concluded, "All my hospital transplantations are
treated as outpatients and return to their homes on the day of operation and
no difficulties of any kind arose there
from."1 In his
synopsis of Robert Jones' life, Mercer Rang said, "Most of all he is
remembered because he was liked and popularity provides opportunities denied
other men."1
What a wonderful tribute to a great man.
The title of this lecture, "The Continuing Development of Shoulder
Replacement," was chosen because I consider shoulder replacements, like
all joint replacements, to be "unfinished business." We have come
a long way, but we certainly have a long journey ahead of us.
Themistocles Gluck, born in 1853 in Jassy, Romania, was the first surgical
pioneer to record a shoulder replacement, performed with use of an ivory
prosthesis, in
18902. However,
perhaps better known is Emile
Pean2. In Paris, in
1893, he implanted a prosthetic shoulder joint for the late effects of
tuberculosis. This prosthesis, which replaced the proximal half of the
humerus, articulated with the scapula through a universal joint. This was a
remarkably sophisticated joint and was certainly very exceptional surgery,
particularly when one considers the state of anesthesia at that time. He was a
friend of Toulouse Lautrec, the great artist who painted a portrait of Emile
Pean performing an operation. Both men were bon viveurs in Paris in the late
1890s. Pean's original prosthetic joint is now on display in the Smithsonian
Institution in Washington, DC.
In the early 1950s, Jackson Burrows at the Royal National Orthopaedic
Hospital in Stanmore, England, started to do massive replacements of the
proximal part of the humerus for the treatment of low-grade tumors and chronic
infections within the proximal end of the
humerus3. The
rotator cuff was usually excised with the tumor, and hence these massive
prostheses, which were fixed by an intramedullary peg into the distal aspect
of the humerus, would subluxate anterosuperiorly under active flexion.
Nevertheless, these prostheses were remarkably long-lasting and reached
high levels of development, eventually replacing the entire humerus with an
articulated shoulder at one end and a prosthetic elbow at the other. Hence,
when the group at the Royal National Orthopaedic Hospital in Stanmore
developed a shoulder replacement to be used in the treatment of arthritis,
they believed that it must be a constrained joint to stabilize the head of the
humerus and prevent subluxation. This constrained joint looked remarkably like
a mini hip replacement at that time with intramedullary cemented fixation and
a captive snap-fit on the glenoid side. Under modern biomechanical principles,
one could predict now that these joints would fail, as the center of rotation
was lateralized and the turning moment about it was so great that loosening
was inevitable.
Loosening and dislocation did occur, and use of the Stanmore prosthesis was
discontinued in the early 1980s. I was involved with a follow-up study of the
prosthesis4 at the
time that it was failing, and I saw many patients who had had the joint
excised. It therefore gave me the opportunity to see the clinical results of
excision arthroplasty around the shoulder following removal of the prosthesis.
Interestingly, a primary excision arthroplasty does very poorly. However, the
secondary excision arthroplasty after the ravages of infection and loosening
with much scarring produced a much better functional result. The shoulders,
although still obviously functioning poorly, had remarkably little pain.
Many other constrained joints were used around Europe in the 1980s, but all
eventually suffered a similar fate of loosening. A common conception at the
time was that the shoulder is a non-weight-bearing joint, as "we don't
walk on our hands"; hence, prostheses were not designed to take the
forces that we now know pass through the shoulder. The science of biomechanics
was advancing rapidly, and in no way can the shoulder be considered a
non-weight-bearing joint when forces greater than body weight pass through the
glenohumeral interface. At that time, I was working at the Royal National
Orthopaedic Hospital in London with Professor Lipmann Kessel who also
developed a constrained joint but with reversed
geometry5, i.e.,
with the ball on the glenoid side and the socket on the humeral side. This
prosthesis was developed to lateralize the center of rotation deliberately and
to attempt to increase the range of motion so that the tuberosities would
clear the acromion on abduction. Again, the problem was lateralization at the
center of rotation, and transmission of force through this bone-prosthetic
interface led to loosening. Use of this prosthesis was also abandoned.
In the 1950s, De Anquin from Argentina produced an acrylic proximal humeral
prosthesis that was nonconstrained and looked rather like the acrylic
prosthesis for the femoral hemiarthroplasty designed by
Judet6. However, the
name associated with shoulder replacement undoubtedly has been that of Charles
Neer, and his shoulder replacement to this day remains the gold standard. He
was working as chief of the fracture service at the Columbian Presbyterian
Hospital in New York in the early 1950s and had a particular interest in the
fracture patterns of the proximal end of the humerus. He described the
four-part classification and realized that some four-part fractures did
particularly poorly because the humeral head became disconnected from its
blood supply and thus later developed osteonecrosis. In the early 1950s,
proximal femoral hemiarthroplasty was being used for the treatment of a
subcapital fracture of the femur, and Neer designed a proximal humeral head
replacement that was fixed by an intramedullary stem to act as a scaffold
around which the proximal end of the humerus was refashioned. In his hands,
this was successful and was a major step
forward7. From the
beginning, Neer stressed that this operation was a reconstruction and, in
particular, a soft-tissue reconstruction. Initially, this was not generally
understood, and others did not achieve such good results, as his advice
concerning meticulous soft-tissue surgery at the proximal end of the humerus
was not heeded. After success in the treatment of fractures, Neer designed a
glenoid component for use in the treatment of arthritis affecting the
glenohumeral joint.
Neer's prosthesis was a monoblock design, and thus many different sizes
were required to fit different patients. The size and cost of stocking these
many prostheses was the initial stimulus to develop modular prostheses such
that different-sized stems could be associated with different-sized heads and
thus a custom-built prosthesis could be made in the operating room with fewer
implants. Roberts et
al.8, in 1991,
demonstrated that the humeral head was offset anteriorly and was not centered
on the humeral shaft; hence, any monoblock design that was not specifically
made for the left or right shoulder could never be considered anatomical.
Boileau and Walch9,
in 1997, described the enormous variation of anatomy at the proximal end of
the humerus, with version, for example, varying between +5° and -55°.
Hence, the so-called third-generation modular shoulder prostheses were
designed to allow for this great variation in an attempt to better mimic
anatomy. Various ingenious multipart modular designs have been marketed so
that head size, neck angle, and offset can be reliably reproduced.
During the 1970s in France, Professor Paul Grammont had been pursuing a
reversed geometry design, which was specifically made for the problem of cuff
arthropathy. With use of nonconstrained joints, the results of prosthetic
replacement in cuff arthropathy were the poorest. Grammont's design placed a
glenosphere on the scapular side and a socket on the humeral side, but the
center of rotation was medialized. Over the years, this has proved to be
better lasting and more successful for these patients who are very difficult
to treat. Interestingly, it has had a great revival of interest over the last
few years. This appeared to be going "back to the future";
however, there are certain very important differences of geometry in
Grammont's design that make it much more favorable.
In the late 1970s, I became interested in the different developmental lines
of shoulder replacement and found that none had been specifically designed for
the problems of arthritis. The historical prostheses designed by Gluck and
Pean had been developed for the late results of infection (tuberculosis); the
original constrained joints in Europe, for the problems of tumor replacement;
Neer's prosthesis, for proximal humeral fracture; and Grammont's design, for
cuff arthropathy. I was gaining experience, reluctantly, in the problems of
the failed shoulder replacement.
If, for instance, a standard cemented humeral stem failed because of
infection or loosening, then a huge loss of bone could be encountered with
major problems when revision surgery was undertaken. The same is true on the
glenoid side. If the component were to loosen, then the cement bolus would
fail in an unpredictable way and again cause a huge loss of bone stock
presenting major reconstruction difficulties. Neer's stem design had become
the accepted standard of fixation, and most joints were fixed on the humeral
side by a cemented stem reaching halfway down the humerus. Later, uncemented
designs were popularized; but, regardless of how it is fixed, a stem leads to
a stress-riser effect at the tip of the prosthesis. Stress-risers have been a
well-recognized problem in the femur for many years, and these fractures can
be a major problem to deal with in both the humerus and the
femur10. There is
no evidence to indicate why the humeral stem should extend half the length of
the humerus, sometimes with the cement tracking down to the elbow in an
attempt to fix the prosthesis. Sometimes, in a polyarthritic patient with
multiple joint replacements, a stem with cement from above may almost meet a
stem from an elbow prosthesis below, causing a major stress-riser effect in
the vacant segment between the two. In the typical elderly patient who is a
candidate for joint replacement, the bone is porotic and hence the chance of
fracture between the two is great.
There are also occasions when a conventional stemmed prosthesis may not be
used if the medullary canal has already been invaded by an intramedullary nail
or screw fixation device or has been distorted by fracture or congenital
malformation.
Another problem of stemmed fixation is the possibility of infection. Once
the medullary cavity is infected, eradicating the infection can be a major
problem requiring a two-stage operation, again with major bone loss and major
reconstruction problems. An additional problem with stemmed fixation is that
of "pilot error." It has been shown that approximately 30% of the
unsatisfactory results after shoulder replacements are due to component
malposition and that, if version is incorrect by >15°, this can lead to
a painful
shoulder11.
Incorrect placement can and does occur, and removal and repositioning of a
stem is a major reconstruction problem. One always has to bear in mind what
would happen if the joint should fail and thus plan a backup or revision
procedure. The long-term failure rate of shoulder replacement is increasing
with time and is directly in line with the failure rates of hip and knee
replacement. Revision arthroplasty is a growing, man-made problem.
Approximately one-third of the operating-room time for arthroplasties is
devoted to revision replacement. Because arthritis affects the shoulder much
less frequently than the hip or the knee, the numbers of shoulder replacements
that are done bear no comparison with the huge numbers of hip and knee
replacements. It has been shown that a large proportion of shoulder
replacements is done by surgeons who perform fewer than two shoulder
replacements a
year12. It has been
shown, and it is reasonable to suppose, that the likelihood of failure can be
inversely related to the frequency of the surgical procedures. It is a fact
that only the best centers report their results, and therefore the overall
failure rate is probably higher than reported, as anybody who works in a
referral center may be aware. This is a major reason for all joint
replacements to be included on a national or international joint registry, and
then the true incidence of revision and failure can be known.
If stems have these problems, why are they still used in great numbers? The
real reason is that, in our surgical training, we learn shoulder replacement
through trauma work, and unquestionably some form of stem is required for
prosthetic fixation when replacing the shoulder because of a fracture.
However, from the standpoint of the treatment of arthritis, the use of a stem
seemed illogical as other methods of fixation could be considered. If we were
to go back to basics and consider the effects of arthritis on a joint, then
the problem has two parts: (1) the effect on soft tissues, which is being
tackled by great medical advances as we can do little to modify this aspect
surgically, and (2) the effects of roughening of the bone end, which we can
address surgically by replacing this surface.
When any prosthetic replacement for arthritis is considered, the only part
of the prosthesis that is of use to the patient is the new, shiny interface
surface. The patient is unaware of whether this surface is fixed by a stem or
cement or by just a pure surface replacement. In the late 1970s, when
investigating surface replacement options, I was unaware, and remained so
until fairly recently, that Zippel in Germany had implanted two surface
replacements that were fixed by a transosseous screw. Steffee and Moore in the
United States were implanting the small hip-resurfacing prosthesis into the
shoulder and, in Sweden, in greater numbers, a surface replacement SCAN
(Scandinavian) cup was being used as a cemented surface
replacement13,14.
During 1978 and 1979, I explored alternative methods of surface replacement
and initially looked at surface replacement of the glenoid with use of glenoid
shells that extended to the undersurface of the acromion. A metal
"mudguard" on the undersurface of the acromion was implanted to
act as an interposition arthroplasty, and later plastic buttons were implanted
on the undersurface of the acromion to act as a subacromial spacer. All of
these avenues of approach had certain design failings and hence were
abandoned. I started to explore surface replacement designs for the humerus
and a nonconstrained glenoid in late 1979 and did extensive anatomical studies
for sizing the proximal end of the humerus and of the blood supply to the
proximal end of the humerus. At that time, surface replacement of the hip was
being popularized and was failing for several different reasons. Surface
replacement was considered rather foolhardy, and the bad experience with the
hip kept surface replacement of the shoulder back for many years.
The biological situation at the proximal end of the humerus is completely
different, and our work at that time gave us every reason to believe that
surface replacement of the shoulder would be successful. I felt that there
were certain important design principles that must be included and that I
would try to develop a prosthesis that was as near to a surface replacement as
feasible with minimal bone removal. It should also incorporate the possibility
of bone-graft impaction and actually making the humeral head better so that
bone loss and cysts could be replaced. Because of the unpredictable method of
failure at the bone-cement interface, I believed strongly that it should be a
cementless fixation. The design had to be simple such that anatomical geometry
could be easily reproduced. If just a surface replacement is used, then
offset, version, and inclination should be automatically accounted for. Many
shoulder replacements are done by surgeons who do very few such
procedures12, and
hence the instrumentation and the surgical principles had to be very simple.
During the early 1980s, many biomechanical studies were made and prototype
testing was introduced between 1984 and 1986. In 1986, I did the first
clinical implants in patients using a surface replacement both on the humeral
side and on the glenoid
side15. On the
glenoid side, an all-high-density polyethylene prosthesis was used with a
finned "magic peg" fixation device that was then in common use for
the tibial fixation of a knee replacement. Eventually, this was changed for
theoretical reasons to a metal-backed glenoid component with an impact-fit,
grooved, tapered peg, which was also used on the humeral
side16. This was
extremely successful, but, like all uncemented prostheses, there was an
incidence of lucent line formation at the bone-metal interface, which was a
worrying feature but did not appear to have any early clinical importance.
Later, osseous ingrowth technology improved enormously, and in 1993 the
prosthesis was hydroxyapatite-coated, which virtually abolished lucent line
formation.
One of the reasons we have to change our approach to joint replacement is
that the patients are changing. Initially, we would only do joint replacement
in very elderly patients who were most disabled and would welcome any return
of function. Now, however, we are asked to do joint replacement in much
younger people. They have an increasing expectation and certainly increasing
use of the prosthesis that they are given. They expect to play contact sports
and live life to the full and see no reason why they should not do so.
Osteonecrosis of the humeral head may be seen after the use of steroids,
and there is an increasing demand for shoulder replacement in young people
because of the use of high-dose steroids in chemotherapy regimes for the
treatment of malignant tumors. The late effects of trauma can occur at any
age, and therefore one has to consider whether the designs we have are capable
of lasting thirty, forty, or fifty years. The straight answer is that we do
not know. We have follow-up data for a maximum of only twenty years. Because
the rate of revision increases with time, if the joint is to fail then it must
fail in a predictable way and in a way that can be resolved. One of the
advantages of a surface replacement is that if it were to fail, revision
surgery is easy. No bone stock is lost, and revision may be done with an
additional surface replacement or a stemmed replacement. If persistent
instability is a problem, the joint is amenable to arthrodesis. Attempting to
perform an arthrodesis on a shoulder after removal of a cemented implant is a
major problem. When a surface replacement is used, bone stock in the head can
be reconstituted with local or allograft bone-grafting.
The indications for surface replacement are exactly the same as those for
any prosthesis used in the treatment of arthritis, and hence surface
replacements can be used in the treatment of osteoarthritis, rheumatoid
arthritis, cuff arthropathy, osteonecrosis, instability, arthropathy, or
postinfective arthropathy. There are limitations to surface replacement. If
there is no humeral head remaining, then a surface replacement cannot be used
and a stemmed prosthesis must be considered. With experience, we have found
that approximately 60% of the original bone stock is required and 40% may be
grafted. Many different types of surface replacement are currently available,
and partial humeral head replacement is available for isolated osteochondral
defects in the head.
We now have results available for surface replacements performed over
nearly twenty years, and the three major groups are patients with
osteoarthritis
(55%)17, rheumatoid
arthritis (21%)18,
and osteonecrosis (9%). The results of shoulder replacement are directly
related to the initial diagnostic group such that, in any reported series of
shoulder replacement, patients with osteoarthritis do best. Patients with
rheumatoid arthritis do not score highly on functional scores, but they score
highly on satisfaction. Patients with cuff arthropathy have historically done
rather poorly, but, with the recent reverse designs, this trend at last seems
to be changing.
One of the continuing debates with shoulder replacement is whether to do a
hemiarthroplasty or a total replacement. The very fact that we continue to
discuss this issue twenty to thirty years later means that the differences are
probably not great. The arguments are very similar to the debate relating to
whether to replace the patella in total knee replacement. Early in our series,
we always attempted to do a total shoulder replacement as, at that time, it
was considered the obvious thing to do—total knee, total hip, and total
shoulder. However, there are times when one cannot do a glenoid replacement
because bone stock is not adequate. Often, this is the case in patients with
rheumatoid arthritis. When we looked at the results of surface replacement
over the years, the results after hemiarthroplasty were standing up very well,
with no statistical difference between hemiarthroplasty and total replacement
procedures. On the basis of this analysis, we have tended more and more to do
hemiarthroplasty. The rationale is that, if plastic is implanted in a joint,
then we are introducing the eventual reason why that joint will fail.
Polyethylene debris causes a macrophage response leading to osteolysis and
possible loosening and failure of the joint. This is a materials problem, and
I am sure that in the future better materials will become available, but at
the moment they are not sufficiently reliable for long-term use for thirty,
forty, or fifty years.
When deciding to do a hemiarthroplasty or a total arthroplasty, one must
balance the possibility of longer-term glenoid wear and loosening in a total
replacement with the possibility of late glenoid erosion in a
hemire-placement, necessitating conversion to a total replacement. When
comparing our results with other directly comparable series of shoulders
managed with a stemmed prosthesis with a similar length of follow-up, the
results of surface hemireplacement appear to be better. In a published series
of shoulders with concentric osteoarthritis, the rate of revision of a
hemireplacement to a total shoulder replacement was approximately 12% at three
years19. Our
surface replacements had a 1% revision rate over four years.
This was also true in patients with rheumatoid
arthritis18. It
appears that the glenoid prosthesis is the problem in the longer term. In a
stemmed prosthesis system, the rate of conversion from hemiarthroplasty to
total replacement is higher than the revision rate for the glenoid component
in a total replacement. The results of surface replacement are substantially
different, and one has to look at the reasons for this. When studying the
geometry of surface replacement, we found that our series of arthritic
patients had a mean medialized erosion of the lateral glenohumeral offset of 6
mm preoperatively, and we managed to restore this by a mean of 6 mm
postoperatively20.
Patients with rheumatoid disease present the most obvious example of what may
be achieved by hemiarthroplasty. If there is, for instance, major
medialization by erosion of approximately 2 cm, then to lateralize the full 2
cm would lead to a completely stiff and painful joint because of the adaptive
shortening of all of the soft tissues. Hence, complete restoration of the
anatomy in some patients may be an unrealizable goal. There must be a
compromise between lateralization to regain reasonable geometry and muscle
tension. It appears that surface replacement can be reliably centered on the
remaining humeral head and not on the shaft, and it can reliably mimic
inclination, retroversion, and posterior offset. The fact that there is no
substantial change in the center of instant rotation suggests that placement
is accurate anatomically.
There are additional differences between stemmed and surface replacement
systems, which could explain the anomaly of different results between
hemiarthroplasty with a stemmed prosthesis and one with surface replacement.
When a surface replacement is done, the version, inclination, offset, and size
are all easily judged and related directly to the anatomy. This is more
difficult to evaluate correctly once the humeral head has been removed in a
hemiarthroplasty with a stemmed prosthesis. Historically, the skills of
shoulder replacement have been learned in the fracture situation, in which the
head has been removed and version has to be guessed during reconstruction.
When a stemmed hemiarthroplasty of a third-generation type is performed,
the thickness of the bone is measured after removal of the head and the new
prosthetic head is placed in the same position as the previously worn head,
i.e., the center of rotation has been medialized. With surface replacement,
the aim is entirely different. When a surface replacement hemiarthroplasty is
performed, the aim is to place the surface such that the center of rotation is
lateralized. Hence, the thickness of the prosthesis has to make up for the
loss of articular cartilage on the glenoid side, loss of bone on the glenoid
side, loss of articular cartilage on the humeral side, and loss of humeral
bone. Thus, when seen in a radiograph, a surface replacement may give the
appearance of being over-stuffed. This is probably because we are used to
looking at straight anteroposterior radiographs and noting the position of the
humeral bone, not the articular cartilage. A surface replacement attempts to
make up this deficit and lateralize the joint at the same time.
In an ideal world with ideal materials, the aim of shoulder replacement
must be to reliably replace both the humeral side and the glenoid side. At the
moment, we can only replace the humeral side reliably in the longer term.
Therefore, until better materials come along, we must address the problem of
the glenoid side. Different methods, such as the interposition of soft tissue,
the subscapularis, and allografts, have been used. For some years, we have
been doing microfracture on the glenoid side, attempting to produce a
fibrocartilaginous bearing surface; second-look arthroscopy certainly suggests
that this is the case.
In the overall series of surface replacements performed in the last twenty
years, the failure rate has been 6.87% throughout all of the diagnostic
categories and design
changes17,18.
However, if we look at the results since the introduction of osseous ingrowth
hydroxyapatite coating for all diagnostic subgroups, the revision rate is
2.6%, which is much more acceptable. If we look at osteoarthritis alone, which
includes 55% of our patients with arthritis, only one replacement, which was
done originally for osteoarthritis, has been revised since 1993 (a revision
rate of 0.71%) and survival of the hydroxyapatite-coated prosthesis was 98% at
ten years.
According to actuarial life-expectancy figures, a patient who is fifty
years old has approximately thirty-one years to live. A forty-year-old
individual has forty years to live. Will our joint replacements last that
long? We truly do not know, but experience leads us to believe that a
substantial number will fail during this period. Therefore, it is necessary
for us to do the minimum to get the maximum, and whatever we do we must have a
fall-back procedure and it must be easily revisable.
In conclusion, surface replacement of the shoulder has been proven to be at
least as successful as stemmed implants with certain advantages for treatment
in young people. The hydroxyapatite coating has been a major advance in
reducing lucent lines and loosening. If complications do occur, then they can
be more easily treated, and the results of surface hemiarthroplasty appear to
be better than stemmed hemiarthroplasty. We have reached a watershed in the
treatment of shoulder arthritis and prosthetic intervention. It is no longer
acceptable to have available only one prosthesis that one knows how to
implant. We must choose the prosthesis for the problem. A stemmed prosthesis
is required for fracture reconstruction. A reverse prosthesis for cuff
arthropathy in the elderly will provide better functional results, and a
surface replacement is required for arthritis. I feel it is no longer
justifiable to continue with intramedullary (either cementless or cemented)
fixation in a straightforward arthritic problem. The geometry and mechanics of
the shoulder joint are now much better understood. The results of shoulder
replacement are better and more reliable. Plastic wear remains a problem.
Long-term glenoid loosening is a problem, and better materials and designs for
the glenoid are required. There is no doubt that we have come a long way, and
we are only just possibly reaching the surface.