Abstract
We examined the pattern of temporal penetration by thirty-two-millimeter-diameter femoral heads into polyethylene liners in a group of 105 hips (103 patients) in which an Arthropor metal-backed cup had been implanted. Each patient was evaluated radiographically and clinically at a minimum of four different postoperative intervals. The initial evaluation was performed a mean of 2.9 weeks (range, one to fifteen weeks) postoperatively, and the latest evaluation was performed a mean of 7.9 years (range, five to ten years) postoperatively. Two-dimensional wear—that is, penetration by the femoral head into the ultra-high molecular weight polyethylene liner—was determined from anteroposterior radiographs of the pelvis with a computer analysis system that calculated the change in the position of the center of the head relative to the center of the cup.Three new findings are reported. First, there was a large difference (mean, 1.1 millimeters) between the center of the head and that of the cup as measured on the initial postoperative radiographs. This difference underscores the need for researchers to consider the initial displacement of the head when measuring and reporting polyethylene wear.Second, although there was wide variation in responses among individuals, temporal examination of the data revealed a trend toward a decreasing rate of penetration with time. Moreover, the rate of penetration appeared to reach a steady-state value after the sixth postoperative year and remained nearly constant until the ninth postoperative year.Third, by comparing the subsets of patients who had the greatest and the least initial penetration by the head, we found that penetration behavior, although remarkably different between the groups in the first three years postoperatively, became similar with time.CLINICAL RELEVANCE: When making decisions regarding individual patients or hip systems that demonstrate penetration by the femoral head into the polyethylene liner, clinicians should consider the patterns of penetration over time. Measurements of the amount and rate of penetration that are based solely on the most recent radiograph do not represent the full clinical picture. We advocate more frequent radiographic follow-up and, when available, analysis of serial radiographs for patients who have excessive penetration by the femoral head into the acetabular liner.
Attempts to measure the wear of ultra-high molecular weight polyethylene components after a total hip arthroplasty usually involve examination of a patient's most recent clinical radiograph. Typically, authors have reported on the cumulative wear process—that is, the total amount of deformation that the component has undergone2,3,6,10,15. From these types of studies, we have developed an understanding of the mean magnitude of wear that polyethylene components undergo, but we have not reached an understanding of how the components arrive at that particular amount of wear. Only a few authors have concentrated on evaluating the wear process over time by studying penetration by the femoral head into the polyethylene liner at annual intervals4,11,16. We employed such a method to study the time-course of two-dimensional penetration by thirty-two-millimeter-diameter femoral heads into a series of polyethylene liners that had been sterilized in ethylene oxide. As deformation of the polyethylene liner is due to both polyethylene creep and the removal of polyethylene particles, measurements of penetration by the femoral head into the liner (more simply referred to as wear) actually represent this combined process of deformation.
The purpose of the present study was to examine the temporal patterns of penetration by the femoral head into the ultra-high molecular weight polyethylene liner in a group of patients who had the same type of acetabular component. By studying a subset of patients who had a high rate of initial penetration and a subset who had a low rate of initial penetration, we hoped to identify early patterns that might be useful in predicting long-term behavior. We also investigated the relationship between penetration by the head and clinical variables such as the age, gender, and weight of the patient and the angle of inclination of the cup.
*One or more of the authors have received or will receive benefits for personal or professional use from a commercial party related directly or indirectly to the subject of this article. No funds were received in support of this study.
†Anderson Orthopaedic Research Institute, 2445 Army Navy Drive, Arlington, Virginia 22206.
Patients chosen for the study were those who had had a primary total hip replacement before 1988 with an Arthropor porous-coated acetabular component (Joint Medical Products, Stamford, Connecticut) that articulated with a thirty-two-millimeter-diameter femoral head. All of the cups had been inserted without cement, and at least one rim screw had been used to lock the polyethylene liner in place. Each patient was evaluated clinically and radiographically at a minimum of four different postoperative intervals. The initial evaluation was performed within the first few weeks postoperatively, and the most recent evaluation was performed five to ten years postoperatively. We evaluated a total of 489 radiographs from 105 hips in 103 patients. There were fifty-eight men and forty-five women. The mean age at the time of the operation was 58.4 years (range, thirty-one to seventy-nine years), and the mean body weight was 77.0 kilograms (range, 40.8 to 129.2 kilograms). The implants had been in situ for a mean of 7.9 years (range, five to ten years) at the time of the most recent evaluation.
All of the polyethylene liners were 4.9 millimeters thick and had been sterilized in ethylene oxide. The metal-backed cups were positioned consistently within the acetabulum at a mean angle of inclination (and standard deviation) of 42 ± 6 degrees. All of the corresponding femoral stems were porous-coated, were inserted without cement, and were manufactured by DePuy (Warsaw, Indiana). One hundred and four of these components were Anatomic Medullary Locking (AML) stems with a femoral offset of 4.0 to 5.1 centimeters, and the remaining stem was a custom-built prosthesis. All of the cups and all but one of the stems were considered stable according to the criteria of Engh et al. Ninety-six of the femoral heads were made of cobalt-chromium alloy, and the remaining nine were made of alumina ceramic.
Penetration by a femoral head into a polyethylene liner occurs as a result of both the removal of polyethylene particles and polyethylene creep. Because the two processes are nearly impossible to distinguish from each other in vivo, references to penetration by the head into the liner actually represent a combination of the wear and creep processes.
We quantified two-dimensional penetration by the head into the polyethylene liner with use of annual anteroposterior radiographs of the pelvis and a specially designed computer analysis system, similar to that developed by Devane et al.5, that consisted of a digitizer tablet and a personal computer running specially designed software. However, unlike the system used by Devane et al., which was able to measure displacement of the head in three dimensions, our system measured displacement in only two dimensions. In other words, we only measured movement of the femoral head in the plane of the anteroposterior radiograph and were not able to assess any additional movement of the head in the plane of the lateral radiograph as Devane et al. did.
The first step in calculating penetration by the head was to align an anteroposterior radiograph on the digitizer tablet so that the interteardrop line was parallel to the edge of the tablet. The computer operator then digitized at least five points around the head of the femoral implant and at least five points around the edge of the cup. The magnification of the head and the distance from the center of the x-ray beam then were used to transform the selected points into circles whose centers represented the center of the femoral head and the center of the metal-backed cup. The distance between the center of the head and that of the cup was used to calculate the amount and direction of penetration by the femoral head into the polyethylene liner. The amount of displacement of the head that was measured on the initial postoperative radiograph was assumed to be the zero position, and penetration by the head into the liner was calculated as subsequent motion from this initial position.
To validate this technique, we machined a known amount of polyethylene from a sample liner5. The precise dimensional change of the inner surface of the liner due to machining was confirmed by measurement with a coordinate measuring machine. A cobalt-chromium femoral head then was secured into the liner, and the cup-head composite was mounted in a specially designed jig to simulate the positioning of an acetabular component as seen on an anteroposterior radiograph. To determine how different positions of the cup might affect the calculations of penetration, anteroposterior radiographs were made with the cup positioned in 5, 15, and 30 degrees of anteversion. Penetration by the head into the liner then was measured five times on each of the three radiographs with use of the computer system. The measured amount of penetration then was compared with the known amount of removed material.
Validation testing revealed that the system tended to underestimate the true amount of penetration by a mean of 0.19 millimeter. Additionally, the magnitude of the computer error was somewhat dependent on the degree of anteversion of the cup. Larger anteversion was related to a slightly smaller error. The mean error was 0.08 millimeter when the sample cup was in 30 degrees of anteversion, 0.22 millimeter when it was in 15 degrees of anteversion, and 0.28 millimeter when it was in 5 degrees of anteversion. We also found that operator error was smaller than computer error. The operator was so consistent in the choice of digitized points that the measurements of penetration were extremely reproducible; the standard deviation of the operator's validation measurements never exceeded 0.11 millimeter. Furthermore, because only one person digitized all of the radiographs, interobserver error was eliminated. Consequently, our method is largely limited only by the accuracy of the computer software.
To investigate differences in patterns of penetration and to identify early patterns that might be useful in predicting long-term behavior, we compared two subsets of patients. The first subset consisted of the twenty patients who had the least penetration by the head into the liner during the first two years postoperatively; these nine men and eleven women were a mean of sixty-one years old at the time of the operation, and they were followed for a mean of 7.2 years. The second subset consisted of the twenty patients who had the most penetration by the head into the liner during the first two years postoperatively; these thirteen men and seven women were a mean of fifty-five years old at the time of the operation, and they were followed for a mean of 8.0 years.
Multiple linear regression analysis was used to determine possible correlations between the total amount of penetration or the rate of penetration and several specific variables: the weight and age of the patient at the time of the operation, the abduction angle of the cup, and the direction of penetration. The Student t test was used to determine if there were any significant differences in the total amount of penetration, the direction of penetration, or the rate of penetration between the men and the women and between the young patients (those who were less than sixty years old) and the elderly patients (those who were sixty years old or more). The Student t test also was used to determine any differences in the rate of penetration between the nine alumina ceramic heads and the ninety-six cobalt-chromium alloy heads to evaluate whether the two types of heads could justifiably be analyzed as a single group.
An unexpected finding in our study was the large difference (mean, 1.1 millimeters; range, 0.4 to 2.5 millimeters) between the center of the head and that of the cup as measured on the initial radiographs, which were made at a mean of 2.9 weeks (range, one to fifteen weeks) postoperatively. However, because penetration by the head was calculated as subsequent displacement from the initial position, this difference is not reflected in the results or in the rates of penetration presented.
At the latest radiographic examination (mean, 7.9 years postoperatively; range, five to ten years postoperatively), the mean penetration by the alumina ceramic femoral heads into the polyethylene liners was the same as the mean penetration by the cobalt-chromium femoral heads into the polyethylene liners (1.3 millimeters). As the ceramic heads had been in situ somewhat longer, the mean rate of penetration by the ceramic heads was slightly, but not significantly, less than that by the cobalt-chromium heads (0.16 compared with 0.17 millimeter per year; p = 0.71). Because the performance of the ceramic heads was not substantially different than that of the cobalt-chromium heads, we combined the two types of heads and examined the results for the group as a whole.
As expected, there was great variation in the amount of penetration among individuals. At the time of the most recent follow-up, migration of the femoral head into the polyethylene liner ranged from 0.2 to 3.8 millimeters (mean and standard deviation, 1.3 ± 0.8 millimeters). The mean angle of penetration was directed 31 ± 40 degrees medial to a vertical line through the center of the head. The rate of penetration also varied widely among individuals. At the time of the latest follow-up, the rate ranged from 0.02 to 0.45 millimeter per year (mean, 0.17 ± 0.09 millimeter per year).
The weight of the patient and the abduction angle of the cup showed no significant relationship with either the amount or the rate of penetration as measured on the latest radiograph (p > 0.55 for all regressions). Men had slightly more penetration (mean, 1.4 millimeters) than women (mean, 1.2 millimeters), but this difference also was not significant (p = 0.18). There was, however, a relationship between the age of the patient and the amount of penetration. Linear regression analysis revealed that increased penetration was associated with a younger age (p < 0.05). This relationship was clearly evident when the population was divided into two age-groups. We found a significant difference between the amount of penetration in patients who were less than sixty years old (mean, 1.5 millimeters) and that in patients who were sixty years old or more (mean, 1.1 millimeters) (p < 0.05). Likewise, the mean rate of penetration for the young patients (0.19 millimeter per year) was higher than that for the elderly patients (0.15 millimeter per year) (p = 0.05).
We found no relationship between the angle of penetration by the head into the liner and the abduction angle of the cup. Interestingly, the angle of penetration seemed to be related to age. The mean angle of penetration was 23 degrees medial to a vertical line through the center of the femoral head for the patients who were less than sixty years old and 38 degrees medial to the vertical line for the patients who were sixty years old or more (p < 0.05). The mean angle was 28 degrees for the men, compared with 34 degrees for the women (p = 0.39).
Temporal examination of the data confirmed the expected trend toward increasing penetration with time (Fig. 1).During the first two years postoperatively, the mean movement of the femoral head was 0.4 millimeter. The mean penetration increased each subsequent year except the sixth year, during which there was slightly less penetration than there had been during the fifth year. However, this decrease may have resulted from the difference in the number of patients for whom five and six-year follow-up radiographs were available. (One hundred and five radiographs were studied initially; sixty-one, at two years; forty, at three years; forty-two, at four years; thirty-eight, at five years; fifty-one, at six years; thirty-five, at seven years; fifty-three, at eight years; forty-one, at nine years; and twenty-three, at ten years.) For the patients who were followed for ten years, the most recent radiograph showed a mean of 1.7 millimeters of penetration.
The rate of penetration by the head into the polyethylene liner was greatest (0.29 millimeter per year) during the first two postoperative years (Fig. 2). The rate decreased in subsequent years and appeared to approach a steady-state value after six years. For example, the mean rate of penetration during the second postoperative year was 93 per cent greater than that during the ninth postoperative year. However, the mean rate in the fifth year was only 53 per cent greater than the rate in the ninth year. In the sixth, seventh, and eighth years, the rate of penetration was within 14 per cent of the rate in the ninth year. Additionally, there was a slight, but not significant (p = 0.17), increase in the rate of penetration from the ninth year (0.15 millimeter per year) to the tenth year (0.18 millimeter per year).
The temporal patterns of penetration in the two subsets of patients who had the greatest and the least initial penetration differed most during the first three postoperative years (Fig. 3). During this period, the rate for the patients who had the greatest initial penetration was at least 215 per cent higher than that for the patients who had the least initial penetration (p < 0.05). During subsequent years, however, the rates of penetration were more similar. Although the rate for the patients who had the greatest initial penetration always remained at least 35 per cent higher than that for the patients who had the least initial penetration, both groups appeared to reach a steady-state rate of penetration by the sixth year.
Polyethylene wear remains a serious problem that limits the longevity of total joint replacements. Osteolysis is an unfavorable biological response to polyethylene wear particles. Additionally, the movement of the femoral head from its original centered position may have negative consequences. For example, gross movement of the femoral head into the polyethylene liner during the wear process produces liner eccentricity that may increase the stress on the liner. Both the biological responses to wear particles and the changes in the biomechanics of the hip system due to such eccentricity make it imperative for researchers to study and to understand the in vivo wear process in order to devise ways to improve the performance of polyethylene. To learn more about the in vivo wear process, we studied the penetration by femoral heads into polyethylene liners at multiple intervals.
An unexpected finding in this study was the 1.1-millimeter difference between the center of the head and the center of the cup as measured on the initial postoperative radiograph. Although this difference was considered to be the zero point for calculations of penetration and therefore was not reflected in the results, such a large difference after a mean of only 2.9 weeks in situ was surprising. We have since learned from the manufacturer8 that this particular cup was designed so that the centers of the head and cup would not be congruent. The cup, which is not a complete hemisphere, was designed so that the center of the head lies at the center of the cup face. Thus, the center of the head initially is offset from the geometric center of the cup. However, we measured initial offsets of as much as 2.5 millimeters and the design of the component does not solely account for such discrepancies. We hypothesize that additional factors may have contributed to the initial difference between the center of the head and that of the cup.
The first possible factor is polyethylene creep. It is probable that the polyethylene liner underwent permanent plastic deformation in the first postoperative month. A recent study by Lee and Pienkowski in which ultra-high molecular weight polyethylene was subjected to a constant physiological load demonstrated that polyethylene creep occurs relatively quickly. Those authors subjected polyethylene to various tests and observed that 85 per cent of creep occurred in the first 1000 minutes of a 10,000-minute test. Virtually all creep occurred within the first 3000 to 4000 minutes of the test, or in less than three days. The tests were performed under continuous loading, a condition that is not physiological. Nevertheless, most of the creep that was produced in those tests occurred rather quickly. On the basis of those findings, we assume that the interval from the operation until the first postoperative radiographic examination (mean, 2.9 weeks) was enough time for some permanent plastic deformation to occur.
The second factor concerns tolerances. With this older generation of modular cups, the tolerance between the ultra-high molecular weight polyethylene liner and the metal backing may not have been precise. Movement of the liner in the plane of the cup face is clearly evident on simple manual manipulation of the liner in this type of cup. Although this motion can be small and difficult to discern on an immediate postoperative radiograph, micromotion of the liner of as much as 181 micrometers has been reported in association with modular cups13. In addition, the tolerance between the outer diameter of the femoral head and the inner diameter of the polyethylene liner may have been imprecise because the components were made by two different manufacturers. Whatever the cause, the initial discrepancy between the center of the femoral head and the center of the cup underscores the importance of considering the initial displacement of the femoral head when measuring and reporting polyethylene wear.
Furthermore, by studying a large group of patients who were followed for five to ten years, we documented a trend toward a decreasing rate of penetration over time. In a similar study, Zichner and Willert examined penetration by the femoral head into Müller components. Those authors described initial rates of penetration of as much as 0.5 millimeter per year, which decreased to 0.1 to 0.2 millimeter per year after five years. The results of our analysis were similar in that the rates of penetration in the first two years after total hip arthroplasty were much higher than those after several years; specifically, the mean initial rate of 0.29 millimeter per year decreased to a minimum rate of 0.15 millimeter per year after nine years.
However, we are careful to avoid direct comparison of our specific penetration values with those reported in other studies for two reasons. First, the goal of our study was to document trends in penetration by the femoral head into the acetabular liner, not to compare our absolute measurements of penetration directly with those previously reported in the literature. Second, our method differs from those previously described. Comparison of absolute numbers calculated with use of different techniques with different levels of accuracy may not be reliable. It is also important to remember that our rates of penetration exclude the penetration that was measured on the initial postoperative radiographs whereas the rates reported in other studies may not.
Additionally, we found that the rate of penetration seemed to stabilize with time and approached a steady-state value after six years. The rate of penetration varied little between the sixth and ninth postoperative years (Fig. 2). It has been reported, however, that the properties of polyethylene deteriorate with time1. Moreover, such deterioration has been linked to the method of sterilization used for the component12,14. Sutula et al. documented that gamma sterilization in air alters the chemical and mechanical properties of polyethylene over time, resulting in high subsurface oxidation, reduced ductility, and reduced strength. It is possible that this deterioration of polyethylene over time could lead to increases in the rate of wear after several years in vivo. A unique aspect of our study, however, is that all of the cups were sterilized in ethylene oxide. According to Sutula et al., polyethylene that has been sterilized in ethylene oxide shows little reduction in mechanical properties over time. This may be one reason why the polyethylene liners in our series showed consistent penetration behavior for as long as nine years postoperatively.
Interestingly, there was a slight increase in the rate of penetration between the ninth and tenth postoperative years. Although this increase might signal the beginning of a period of rapid penetration, only twenty-three patients had more than nine years of follow-up and this subgroup may not be representative of the entire population.
When we compared the temporal wear patterns in the patients who had the most initial penetration with those in the patients who had the least initial penetration, we observed a large difference between the two groups during the first three years postoperatively. The rate of wear was considerably higher for the patients who had the most initial penetration than for those who had the least initial penetration, and there was much greater variation between individual patterns. However, the differences in penetration behavior between the two groups became smaller and the individual patterns became less varied as time progressed. This trend is evidenced by the smaller standard deviations with increasing time. Finally, we found that the groups had similar patterns of penetration between four and nine years postoperatively, having reached a steady-state rate of penetration. During these years, the rate for the patients who had the most initial penetration remained at least 35 per cent higher than that for the patients who had the least initial penetration (Fig. 3). However, the rates for the former group incorporate the very high two-year rate of penetration. If we subtract the displacement that the heads underwent in the first two years postoperatively by assigning the position of the head at two years as our starting point, we find that the rates of penetration for both groups are similar. Specifically, from four to nine years postoperatively, incremental movement of the head into the polyethylene liner for both groups was nearly identical.
This finding demonstrates that assumptions about long-term behavior that are based on the initial movement of the head may be inaccurate. Moreover, it emphasizes that patients who have rapid penetration by the head into the liner need to be followed more frequently so that better assessments of wear and appropriate decisions regarding treatment can be made. We caution, however, that the patterns of penetration in such patients should be observed over time only if the polyethylene liner is of adequate thickness. If temporal examinations reveal that the rate of penetration stabilizes over time, an early revision may be avoided. However, if subsequent evaluations reveal that penetration is continuing at a high rate, or if the polyethylene is not thick enough for the further development of penetration patterns to be observed safely, revision should be considered.
When serial radiographs are not available, the clinician must be aware of penetration by the head into the liner in the first weeks postoperatively as well as the increased rate of penetration during the first three years postoperatively (Fig. 4). We believe that measurements of the amount and rate of penetration that are based solely on the most recent clinical radiograph do not represent the complete clinical picture. Clinical decisions about individual patients and judgments about particular hip systems should be based on patterns of penetration over time, not just on the absolute magnitude of penetration.
Finally, the measurement of movement of the femoral head in only two dimensions has limitations. There may be a substantial amount of movement in the plane of the lateral radiograph that our technique prevents us from evaluating. A three-dimensional analysis that includes displacement of the femoral head in the lateral plane may yield different results.
NOTE: The authors thank William Moore for assistance with the organization and collection of the data and DePuy (Warsaw, Indiana) for software development.
Bostrom, M. P.; Bennett, A. P.; Rimnac, C. M.; and Wright, T. M.: The natural history of ultra high molecular weight polyethylene. Clin. Orthop.,309: 20-28, 1994.30920
1994
[PubMed]
Callaghan, J. J.; Pedersen, D. R.; Olejniczak, J. P.; Goetz, D. D.; and Johnston, R. C.: Radiographic measurement of wear in 5 cohorts of patients observed for 5 to 22 years. Clin. Orthop.,317: 14-18, 1995.31714
1995
[PubMed]
Cates, H. E.; Faris, P. M.; Keating, E. M.; and Ritter, M. A.: Polyethylene wear in cemented metal-backed acetabular cups. J. Bone and Joint Surg.,75-B(2): 249-253, 1993.75-B(2)249
1993
Charnley, J., and Halley, D. K.: Rate of wear in total hip replacement. Clin. Orthop.,112: 170-179, 1975.112170
1975
[PubMed]
Devane, P. A.; Bourne, R. B.; Rorabeck, C. H.; Hardie, R. M.; and Horne, J. G.: Measurement of polyethylene wear in metal-backed cups. I. Three-dimensional technique. Clin. Orthop.,319: 303-316, 1995.319303
1995
[PubMed]
Devane, P. A.; Bourne, R. B.; Rorabeck, C. H.; MacDonald, S.; and Robinson, E. J.: Measurement of polyethylene wear in metal-backed cups. II. Clinical application. Clin. Orthop.,319: 317-326, 1995.319317
1995
[PubMed]
Engh, C. A.; Massin, P.; and Suthers, K. E.: Roentgenographic assessment of the biologic fixation of porous-surfaced femoral components. Clin. Orthop.,257: 107-128, 1990.257107
1990
[PubMed]
Joint Medical Products: Personal communication, 1996.
Lee, K.-Y., and Pienkowski, D.: Compressive creep of UHMWPE varies with time, pressure, and radius of extruded rod stock. Trans. Orthop. Res. Soc,21: 468, 1996.21468
1996
Livermore, J.; Ilstrup, D.; and Morrey, B.: Effect of femoral head size on wear of the polyethylene acetabular component. J. Bone and Joint Surg,72-A: 518-528, April 1990.72-A518
1990
Rimnac, C. M.; Klein, R. W.; Betts, F.; and Wright, T. M.: Post-irradiation aging of ultra-high molecular weight polyethylene. J. Bone and Joint Surg,76-A: 1052-1056, July 1994.76-A1052
1994
Rimnac, C. M.; Wilson, P. D., Jr.; Fuchs, M. D.; and Wright, T. M.: Acetabular cup wear in total hip arthroplasty. Orthop. Clin. North America,19: 631-636, 1988.19631
1988
Smith, S. E.; Fehring, T. K.; Braun, E. R.; Mobley, C. E., Jr.; and Wang, P.-L.: Liner motion in modular acetabular components. Read at the Annual Meeting of The American Academy of Orthopaedic Surgeons, Atlanta, Georgia, Feb. 22, 1996.
Sutula, L. C.; Collier, J. P.; Saum, K. A.; Currier, B. H.; Currier, J. H.; Sanford, W. M.; Mayor, M. B.; Wooding, R. E.; Sperling, D. K.; Williams, I. R.; Kasprzak, D. J.; and Surprenant, V. A.: Impact of gamma sterilization on clinical performance of polyethylene in the hip. Clin. Orthop.,319: 28-40, 1995.31928
1995
[PubMed]
Woolson, S. T., and Murphy, M. G.: Wear of the polyethylene of Harris-Galante acetabular components inserted without cement. J. Bone and Joint Surg,77-A: 1311-1314, Sept. 1995.77-A1311
1995
Zichner, L. P.,, and Willert, H.-G.: Comparison of alumina-polyethylene and metal-polyethylene in clinical trials. Clin. Orthop.,282: 86-94, 1992.28286
1992
[PubMed]