A major challenge for the surgeon performing hip arthroplasty is to provide a durable interlock between the implant and the bone. While cement can provide lasting fixation of femoral stems1,35-37, reports of high rates of failure in patients who are younger (fifty years old or less) and more active6,8,14 have stimulated the exploration of fixation of the femoral component without cement as an alternative2,5,7,11,12,16,20,24,28. Although porous coatings, which allow an interlock between the implant and the bone, remain the primary means of fixation without cement, a newer method of fixation of femoral stems with use of crystalline hydroxyapatite has been reported11,12,21,22. Application of a thin but dense fifty-micrometer-thick layer of hydroxyapatite by plasma-spraying onto a roughened titanium surface has resulted in rapid and predictable ongrowth of bone in experiments involving dogs and in clinical studies involving humans10-12,21-23.
The transfer of load from the prosthesis to the bone, an important function of a femoral component, alters the pattern of stress transfer in the femur compared with the normal state without an implant. The stresses that occur in the bone after implantation of the stem are due to a combination of axial, bending, and torsional loads29,30. The location and magnitude of stress transfer from the femoral component to the bone vary according to the shape and stiffness of the implant15,17,18,29,30, the level of the coating for fixation15,29,30, the stiffness of the bone preoperatively and the resultant hip-force loads after implantation15,18,19,29,30, and the degree of bonding of the implant to the bone29,30. Femoral implants that are designed to be inserted without cement are large and fill the intramedullary canal; they have a greater modulus of elasticity and, therefore, of stiffness than does bone. Assuming that the bone bonds to the coating of the implant, the level of the coating (proximal or more extensive) will then influence the degree of proximal stress-shielding and the amount of distal stress transfer. Therefore, for a given type of bone and level of activity, adaptive bone-remodeling varies according to the different design characteristics of each type of femoral implant. Serial radiographs of femoral implants inserted without cement have been a reliable indicator of the fixation and stability of the implant, adaptive bone-remodeling, and osteolysis3,5,11,17,18,28.
The purpose of the current report was to describe the radiographic changes, at a minimum of fifty-eight months after implantation, around titanium-alloy femoral stems that were proximally coated with hydroxyapatite. Recognition of these extensive and predictable changes will aid in the understanding of the osseous remodeling that occurs about such well bonded proximally coated femoral implants.
†M. H. & D. Orthopedic Associates, 725 Cherrington Parkway, Suite 200, Moon Township, Pennsylvania 15108.
‡Department of Orthopaedic Surgery, Indiana University School of Medicine, 541 Clinical Drive, Room 600, Indianapolis, Indiana 46202.
§744 Paiute Place, Franklin Lakes, New Jersey 07417.
As part of a multicenter clinical trial, eighteen surgeons at fifteen sites implanted 436 hydroxyapatite-coated femoral stems during primary total hip arthroplasties that were performed between January 1988 and November 1990. A subgroup of 251 femoral stems were implanted in 227 patients by sixteen surgeons before July 1989, and the patients included in the current report were from that population. Exclusions from the database included seven patients (eight hips) who had died, ten patients (ten hips) who had been lost to follow-up, and nine patients (nine hips) who had had revision of the femoral stem before the five-year follow-up evaluation. Of the remaining 201 patients (224 hips), 156 were followed for six years and twenty-four were followed for seven years. The mean duration of follow-up was seventy-one months (range, fifty-eight to eighty-seven months).
All patients had insertion of a grit-blasted, collarless, straight titanium-alloy femoral implant (Osteonics, Allendale, New Jersey) with normalizations and a dense fifty-micrometer-thick layer of hydroxyapatite applied circumferentially by plasma-spraying to the proximal one-third of the stem (Fig. 1). Normalizations are steps, machined into the proximal anterior and posterior surfaces, that are intended to convert shear to compressive stress at the bone-stem interface. The proximal coating contained more than 90 per cent hydroxyapatite by weight and has been described previously21-23.
One hundred and thirty-eight of the 251 stems were implanted in men and 113, in women. The mean age of the patients at the time of the operation was fifty years (range, eighteen to seventy-seven years). The mean weight was seventy-eight kilograms (range, thirty-six to 136 kilograms). The mean weight of the women was sixty-nine kilograms (range, thirty-six to 136 kilograms), and the mean weight of the men was eighty-six kilograms (range, fifty-two to 120 kilograms). The most common diagnoses were osteoarthrosis (158 hips; 63 per cent), avascular necrosis (forty-nine hips; 20 per cent), rheumatoid arthritis (sixteen hips; 6 per cent), and post-traumatic osteoarthrosis (fourteen hips; 6 per cent).
Radiographic and clinical data were collected preoperatively, early postoperatively (at six to eight weeks), at six months, at one year, and yearly thereafter. The level of pain, as well as functional parameters such as the distance that the patient could walk, stair-climbing, the need for external support, sitting ability, limp, range of motion, and participation in recreational activities, were evaluated at each visit, and a composite Harris hip score27 was calculated. A score of 90 to 100 points is considered an excellent result; 80 to 89 points, a good result; 70 to 79 points, a fair result; and 69 points or less, a poor result. Anteroposterior and lateral radiographs of the hip were made preoperatively and at all postoperative visits. The preoperative radiographs were evaluated for the type of bone, according to the classification of Dorr13. The postoperative radiographs were evaluated for radiolucency in the zones described by Gruen et al.26, periosteal cortical hypertrophy, cancellous condensation, atrophy of the calcar of the femur, subsidence of the implant, erosion of cortical bone (osteolysis), and heterotopic bone formation according to the classification of Brooker et al.4. Zones 1 and 7 on the anteroposterior radiographs and zones 8 and 14 on the lateral radiographs represent the proximal hydroxyapatite-coated areas of the femoral stem.
Radiolucency was defined as a radiolucent line that was parallel with and in close proximity to the implant and was associated with a thin radiopaque layer of bone paralleling the line. A radiolucent line must encompass at least 50 per cent of the length of the zone to be recorded according to the criteria of Gruen et al.26. Periosteal cortical hypertrophy was recorded if there was an increase in the outside diameter of the cortex in each zone about the implant. Cancellous condensation was defined as new-bone formation between the implant and the endosteal surface of the femur as seen on follow-up radiographs compared with the appearance on the radiographs made immediately postoperatively. Such condensation was also classified according to the zones of Gruen et al.26. Atrophy of the calcar was recorded as a loss of either height or width of bone. Subsidence was measured from the tip of the greater trochanter to the shoulder of the implant. Osteolysis, defined as an erosive lesion anywhere about the femoral implant, was classified according to zone26.
In addition to revision of the stem, we used two other indices to evaluate the performance of the implant. The rate of mechanical failure was based on the number of stems that had been revised because of aseptic loosening as well as the number that were in situ and were determined to be loose according to radiographic criteria. The rate of clinical failure was based on the number of well fixed stems that had been revised because of osteolysis or pain as well as the number associated with pain that limited activity. The combined rate of failure was the sum of the mechanical failures and the clinical failures.
The mean Harris hip score was 48 points (range, 19 to 77 points) preoperatively, 95 points (range, 30 to 100) points at one year, and 95 points (range, 48 to 100 points) at the most recent follow-up examination. At the time of the most recent follow-up, 208 hips (93 per cent; 190 patients) had a good or excellent result. Two hundred and seven hips (92 per cent; 185 patients) were pain-free, and seventeen hips (8 per cent; sixteen patients) were associated with mild or marked pain. Of these sixteen patients, seven (eight hips) had pain in the groin and a radiographically unstable acetabular component; one (one hip), in whom the femoral stem was radiographically stable, had moderate pain in the thigh; and eight (eight hips) had mild activity-related pain. When pain in the groin and the buttocks was eliminated from the calculation, only four patients (four hips; 2 per cent) had mild or moderate activity-related pain in the thigh.
Other functional results included no limitation with regard to the distance that could be walked (190 hips; 85 per cent), no limp or a mild limp (213 hips; 95 per cent), the ability to climb stairs foot over foot (212 hips; 95 per cent), and the ability to sit in any chair (217 hips; 97 per cent).
In the entire population of 251 hips, complications included nineteen intraoperative undisplaced fractures of the femoral neck (8 per cent), six postoperative femoral fractures (2 per cent), a deep joint infection in two hips (1 per cent), aseptic loosening of the femoral stem in two hips (1 per cent), and a dislocation in fifteen hips (6 per cent). Of the nineteen intraoperative undisplaced fractures of the femoral neck, ten received no treatment, seven were treated with cerclage wire, one was treated with a number-5 Ethibond suture (Ethicon, Somerville, New Jersey), and one was treated with a tension wire. None of these fractures led to subsidence or failure of the implant. Of the six postoperative fractures, four were treated with open reduction and internal fixation, one led to revision of the femoral component, and one was not treated. Of the fifteen dislocations in thirteen patients, eight (six patients) were reported by one surgeon; four, by two surgeons (two by each); and three, by three surgeons. A posterior operative approach had been used in all of these patients.
There were nine revisions of the femoral component: two were performed for a deep joint infection; one, for a postoperative fracture; three, for pain in the hip or the thigh (all components in these three hips were well fixed and were difficult to remove); one, at the time of revision of the acetabular component although the femoral component was not loose; and two, for aseptic loosening. Thus, the rate of revision for aseptic loosening was 1 per cent (two implants). Both revisions of an aseptically loose femoral implant were performed during the first two years of the study. One of these revisions was in a patient who had a subtrochanteric non-union as the result of an osteotomy performed for an angular deformity at the time of the index operation. The other revision was in a patient who had not had pain until she had increased her level of activity eight months after the index operation; the pain persisted, and the revision was performed two years after the index operation. The three revisions done for pain in the hip or the thigh and the one revision done at the time of a revision of the acetabular component were performed by surgeons who did not participate in the study. The surgeon who did the revision at the time of the acetabular procedure interpreted the distal reactive lines around the uncoated portion of the stem as a sign of loosening31. However, the radiographs revealed evidence of an osseointegrated stem, with circumferential cancellous condensation at the middle portion of the stem. At the time of the revision, the implant was found to be tightly fixed and osseointegrated, and the stem was removed with great difficulty31.
The rate of aseptic loosening was 1 per cent (two) of the total population of 251 hips. No stem was radiographically loose; therefore, the rate of mechanical failure was also 1 per cent (two hips). The combined rate of failure (2 per cent; six hips) includes two stems that were revised for aseptic loosening, no stems that were loose radiographically, three well fixed stems that were revised for pain in the hip or the thigh, and one radiographically stable stem that was associated with activity-limiting pain.
The radiographic evaluation (Figs. 2-A, 2-B, 2-C, 2-D, 2-E through 2-F) of the 224 hips revealed that radiodense lines during the first two postoperative years occurred most often (in 170 hips; 76 per cent) around the uncoated distal tip of the stem; this rate had decreased, at the time of the most recent evaluation, to 68 per cent (152 hips). These radiolucent lines were always parallel, never divergent, and never associated with formation of a so-called pedestal. Analysis of the proximal zones with the hydroxyapatite coating as seen on the anteroposterior and lateral radiographs revealed that four hips (2 per cent) had a radiolucent line in zone 8 and one hip (0.4 per cent) had one in zone 1, at one year and at the time of the most recent follow-up. At one year, cancellous condensation was noted most often (in 141 hips; 63 per cent) at the distal end of the hydroxyapatite coating (between zones 6 and 7) on the medial side of the component on the anteroposterior radiograph. Progressive new-bone formation between the implant and the endosteum occurred in all zones from one through five years or more. The most dramatic occurrences were in the middle and distal regions of the stem, involving a maximum of 206 hips (92 per cent) medially in zone 6, 190 hips (85 per cent) laterally in zone 2, and 181 hips (81 per cent) posteriorly in zone 13. This new-bone formation in the middle portion of the stem began at the distal end of the hydroxyapatite coating and progressed both distally and proximally. Substantial increases in the prevalence of condensation were also noted around the distal portion of the stem (zones 3, 5, 10, and 12); notably, there was an increase, between the one-year and most recent follow-up periods, from nine to 134 hips (from 4 to 60 per cent) that had condensation in zone 5 and from thirteen to 152 hips (from 6 to 68 per cent) that had it in zone 12. Periosteal cortical hypertrophy was noted most frequently in zone 5 (distal medial) on the anteroposterior radiographs and had occurred in twenty-four hips (11 per cent) by one year; this rate had increased progressively, to 105 hips (47 per cent) at the most recent follow-up examination (Fig. 2-D). The prevalence of hypertrophy in some zones about the middle and distal portions of the stem increased fourfold to sixfold during the follow-up period. At the most recent examination, forty-three hips (19 per cent) had hypertrophy in zone 3; forty-three (19 per cent), in zone 6; and twenty-nine (13 per cent), in zone 2. Hypertrophy was not commonly seen on the lateral radiographs, but when it was seen it was most often found in zone 12 (thirteen hips; 6 per cent) at five years or more.
Another radiographic finding at the time of the most recent follow-up was subsidence of more than three millimeters in three hips. In one of these hips, subsidence noted at three years had not progressed at five years. In the other two hips, the subsidence was associated with immediate postoperative trauma, and it did not progress or lead to failure of the stem after six and twelve-week periods of protected weight-bearing. Atrophy of the calcar of the femur (loss of either height or width of the bone) was noted in 141 hips (63 per cent). Erosive osteolytic lesions at the level of the femoral neck resection were found in ninety-six hips (43 per cent) at the most recent follow-up examination. These lesions were associated with an age of less than fifty years, use of a thirty-two-millimeter-diameter femoral head, and use of a liner with a thickness of less than eight millimeters.
Proximal erosion of cortical bone was seen in fifty-four (54 per cent) of 100 hips in patients who were less than fifty years old, compared with forty-two (37 per cent) of 114 hips in patients who were fifty years old or more (p = 0.009); in thirty-nine (62 per cent) of sixty-three hips that had a thirty-two-millimeter-diameter femoral head, compared with fifty-seven (37 per cent) of 154 hips that had a femoral head with a smaller diameter (p = 0.0008); and in thirty-seven (60 per cent) of sixty-two hips that had a liner that was less than eight millimeters thick, compared with fifty-nine (38 per cent) of 155 hips that had a liner with a thickness of eight millimeters or more (p = 0.003). (A few hips were excluded from each analysis because the radiographs were inadequate or a bipolar prothesis had been used.)
Only one hip (0.4 per cent) had intramedullary (distal) osteolysis. The osteolytic area was less than five millimeters in its greatest dimension; it was seen at five years in zone 9 on the lateral radiograph, and it had not changed or progressed by the sixth year of follow-up. According to the criteria of Engh et al.18, all of the stems were considered to have osseous stability by two years and remained stable throughout the follow-up period. There was no deterioration at the bone-prosthesis interface that would indicate loss of stability or fixation of any stem.
Heterotopic bone formation was found in eighty-three hips (37 per cent); in two hips (1 per cent), the heterotopic bone was classified as grade 4 according to the system of Brooker et al.4. No patient who had heterotopic ossification had clinical symptoms or limitations related to this radiographic finding, and it did not interfere with the measurement of subsidence from the tip of the greater trochanter.
The initial bone quality could be evaluated in 220 hips. According to the classification of Dorr13, seventy-eight hips (35 per cent) had type-A bone, 131 (60 per cent) had type-B, and eleven (5 per cent) had type-C. When the radiographic changes were evaluated on the basis of the type of bone, the only significant findings, with the numbers available, were those with regard to radiolucency and cortical hypertrophy. The rate of zone-5 radiolucency associated with type-C bone (four of eleven hips) was higher than that associated with type-B bone (eleven [8 per cent] of 131 hips; p = 0.03) or type-A bone (three [4 per cent] of seventy-eight hips; p = 0.003). The rate of zone-10 radiolucency associated with type-B bone (fifty-seven [44 per cent] of 131 hips) was higher than that associated with type-A bone (twenty-two [28 per cent] of seventy-eight hips; p = 0.02). The rate of zone-12 radiolucency associated with type-C bone (six of eleven hips) was higher than that associated with type-A bone (thirteen [17 per cent] of seventy-eight hips; p = 0.01). The rate of hypertrophy associated with type-C bone (nine of eleven hips) was higher than that associated with type-A bone (thirty-three [42 per cent] of seventy-eight hips; p = 0.01). The rate of hypertrophy of type-B bone (seventy-six [58 per cent] of 131 hips) was higher than that associated with type-A bone (thirty-three [42 per cent] of seventy-eight hips; p = 0.02). With the numbers available, no significant differences were found, among the types of bone, with regard to atrophy of the calcar, cancellous condensation, or proximal erosion.
Proximal (metaphyseal) stress-shielding was believed to be present about many of the femoral stems at the most recent follow-up examination; however, we made no effort to quantitate this finding, as it was difficult to identify clearly on routine radiographs because of both variations in the radiographic technique and the contrast in density that occurs between zones adjacent to the hydroxyapatite-coated part of the stem as compared with those adjacent to the uncoated part. The zones adjacent to the uncoated distal part of the stem routinely had increased density with cancellous condensation at the one through five-year follow-up evaluations.
In the 1980's, it was demonstrated that osseous fixation of an implant without cement can be achieved when a well designed implant is press-fit into the femur3,5,16,17,20,21,28. On the basis of that experience, we have a better understanding of the adaptive bone-remodeling that occurs about proximally and extensively coated porous femoral stems15,18,19,22,28. Our current clinical experience (mean duration of follow-up, seventy-one months) with a hydroxyapatite proximally coated titanium-alloy implant has demonstrated predictable early and progressive bone-remodeling about the femoral stem. We previously reported the early (two and three-year) clinical and radiographic results of use of this stem11,12. The radiographic changes seen at five years or more are consistent with those earlier results. The purpose of the current report was to describe the radiographic changes around titanium-alloy femoral stems that had been proximally coated with hydroxyapatite. Recognition of these extensive and predictable changes will aid in the understanding of the bone-remodeling that occurs about such well bonded proximally coated femoral implants.
Finite element analysis has been used to estimate theoretically the differences in load-transfer mechanisms and in stress patterns among cemented, extensively coated, and proximally coated femoral stems in total hip arthroplasty29,30. Those studies suggested that the stress patterns in all configurations of bonded stems, whether the stem had been inserted with or without cement, were determined by the bending forces about the hip and were qualitatively similar; differences occurred primarily because of the size and rigidity of the stems implanted without cement. Such stems, by virtue of their increased filling of the canal and, thus, their structural stiffness, create more cortical stress-shielding, more distal stress transfer, and less proximal stress transfer. While some degree of stress-shielding in the proximal part of the femur is inevitable with any design of stem, cemented stems are associated with less relative proximal stress-shielding than are stems implanted without cement. Compared with fully coated stems, proximally coated stems, which by design are wedged into the proximal part of the femur, are associated with a more uniform stress transfer over the full length of the stem28,29.
The addition of a hydroxyapatite proximal coating results in early and consistent bone-to-stem apposition and thus facilitates load transfer from the implant to the bone. This mechanism of load transfer appeared to be substantiated on the radiographs that were made over a follow-up period that averaged six years. The radiographic manifestation of such load transfer in the form of either an increased or a more slowly decreasing density of proximal bone has been verified from analysis of specimens obtained intraoperatively and at autopsy, as well as from ongoing dual-energy x-ray absorptiometry studies9,32-34. The results of these retrieval analyses and dual-energy x-ray absorptiometry studies were consistent with the theoretical estimations made with finite element analysis.
The radiographically evident bone-remodeling observed about femoral stems that are proximally coated with hydroxyapatite constitutes evidence of extensive progressive adaptive remodeling of femoral bone that has not, to our knowledge, been reported previously in association with proximally coated porous femoral implants5,15,28. We believe that this reflects a higher degree of early and rigid proximal bone fixation to the hydroxyapatite coating. According to the radiographic criteria of Engh et al.18, the only major signs of osseointegration are the absence of complete reactive lines adjacent to the ingrowth or ongrowth surface of an implant and the presence of so-called spot welds of endosteal new bone contacting the ongrowth surface. The only major sign of stability of the implant is the absence of migration18. Progressive subsidence of the hydroxyapatite-coated stems was not seen in this study. No hip had circumferential radiolucency, and only four hips (2 per cent) had isolated radiolucency in zone 8. There was cancellous condensation (spot welds) at the junction of the coated and uncoated portions of all of the stems. No change or deterioration in the proximal bone-prosthesis interface was noted during the follow-up period. Reactive lines, commonly seen around the uncoated distal portion of the stem, were always parallel, were never associated with formation of a so-called pedestal, and tended to decrease in the zones about the distal portion of the stem as cancellous condensation increased progressively during the follow-up period. We believe that these reactive lines may indicate micromotion between the uncoated, unfixed distal portion of the stem and the femoral diaphysis because of the initial lack of bonding and the mismatch in the structural stiffness between the stiffer titanium stem and the bone.
Patients who had type-B or type-C bone had a higher prevalence of radiolucency around the uncoated distal tip compared with those who had type-A bone. We believe that this was due to the less intimate distal endosteal contact between the distal portion of the stem and the host bone and to a greater mismatch in stiffness between the bone and the stem in these patients. Cancellous condensation was most often noted at the junction of the coated and uncoated portions of the stem. It was seen medially in 141 hips (63 per cent) at one year and in 206 hips (92 per cent) at the most recent follow-up examinations. It was seen at some point around the stem at the distal end of the hydroxyapatite coating in all hips between the two-year and most recent follow-up examination. We consider the progressive increase in cancellous condensation throughout the zones adjacent to the middle and distal portions of the stem to be due to bone-remodeling in response to the stress transfer at the middle portion of the stem, as predicted by finite element analysis29,30. We believe that this progressive condensation provides secondary stabilization of the femoral implant. Furthermore, these changes indicate a high degree of stress transmission through a proximally well fixed implant and are also associated with atrophy of the calcar (141 hips; 63 per cent) and with the metaphyseal stress-shielding that was seen in many hips. The prevalence of cancellous condensation did not differ among bone types A, B, and C. The prevalence of periosteal cortical hypertrophy throughout the zones adjacent to the middle and distal portions of the stem was significantly higher in type-B and type-C bone than in type-A bone (p = 0.02 and 0.01). These findings suggest a relatively greater transfer of stress at the middle and distal portions of the stem in femora with poorer initial quality of the bone. In addition, the very low rate of hypertrophy in the anteroposterior plane (zones 8 through 14 on the lateral radiographs) as compared with the rate in the mediolateral plane (zones 1 through 7 on the anteroposterior radiographs) indicates a mid-diaphyseal load transfer of mediolateral bending forces, which is in accordance with the assumptions made in the finite element analysis as described earlier. Finite element analysis, analysis of specimens retrieved at autopsy, and clinical studies of extensively coated stems have shown that the initial quality of the bone plays an important role in stress-remodeling around femoral implants inserted without cement15,17,29,30,34. Femora with better-quality bone show less adaptive bone-remodeling after implantation of the stem than do those with poorer-quality bone and the same size of implant.
Osteolysis is a major concern after total hip arthroplasty38. Recent reports of cemented femoral stems have documented rates of intramedullary osteolysis ranging from 0 per cent (of forty-one hips; thirty-eight patients25) to 6 per cent (of 100 hips; eighty-one patients35) at a minimum of five years postoperatively. The rate of such osteolysis for stems inserted without cement has ranged from 1 per cent (of 217 hips; 217 patients16) to 7 per cent (of 100 hips; 100 patients28) to 29 per cent (of forty-one hips; thirty-nine patients25). The 0.4 per cent rate of non-progressive osteolysis (one hip) in the current study strongly suggests that a circumferential hydroxyapatite coating is an effective mechanism for sealing the femur from migration of polyethylene particles. The absence of osteolysis is also evidence that hydroxyapatite particles are not osteolytic. The erosive lesions found at the level of the femoral neck resection were presumably secondary to polyethylene wear debris and were associated with an age of less than fifty years (p = 0.009), the use of a thirty-two-millimeter-diameter femoral head (p = 0.0008), and the use of a liner with a thickness of less than eight millimeters (p = 0.003).
In summary, the findings of radiographic evaluation of femoral stems that were proximally coated with hydroxyapatite suggest a high rate of early and continuing fixation with predictable and consistent adaptive bone-remodeling within the femur. These radiographic findings corresponded to a good or excellent clinical result in 208 hips (93 per cent; 190 patients) at five years or more, to early relief of pain, and to low rates of activity-related pain in the thigh (2 per cent; four hips) and mechanical failure (1 per cent; two hips). According to the radiographic criteria for stability and fixation described by Engh et al.18, the absence of circumferential radiolucency and subsidence and the presence of so-called spot welds indicated that all of the femoral stems in this series were osseointegrated at two to six years.
NOTE: The authors acknowledge the following investigators who contributed to the study: Benjamin Bierbaum, M.D., Boston, Massachusetts; John Cardea, M.D., Richmond, Virginia; Michael Christie, M.D., Nashville, Tennessee; Omar Crothers, M.D., Portland, Maine; Joseph Dimon, III, M.D., Atlanta, Georgia; Vincent Eilers, M.D., St. Paul, Minnesota; William Jaffe, M.D., New York, N.Y.; Randall Lewis, M.D., Washington, D.C.; James Lindberg, M.D., Denver, Colorado; David Mattingly, M.D., Boston, Massachusetts; William Stillwell, M.D., Smithtown, New York; James Turner, M.D., Cedar Rapids, Iowa; Anthony Unger, M.D., Washington, D.C.; and Richard Zimmerman, M.D., Portland, Oregon. They also thank Anne Serekian for her assistance in the preparation of this manuscript.