Background: Despite the expanding body of literature on radial head arthroplasty, the increasing understanding of elbow anatomy, biomechanics, and kinetics, and the evolution of surgical techniques and prosthesis designs, there is currently no evidence to support one type of radial head prosthesis over another. The purposes of the present report were to review the literature and to explore the association between prosthesis design variables and the timing of surgery and the outcome of modern radial head arthroplasty.
Methods: The literature search was limited to studies involving skeletally mature patients. Major databases were searched from January 1940 to May 2015 to identify studies relating to functional and subjective outcomes and radiographic results after radial head arthroplasty.
Results: Thirty articles involving 727 patients were included. Seventy percent of the implants were made of cobalt-chromium, 15% were made of pyrocarbon, 9% were made of titanium, and 6% were made of Vitallium. Seventy percent were monopolar, and 30% were bipolar. Twenty-one percent were cemented in place, 32% were press-fit, 32% were intentionally loose-fit, and 15% were fixed with an expandable stem. The weighted average duration of follow-up was 45 months. The rate of revision ranged from 0% to 29% among studies. The incidence of revision was 8% during 2,714 person-years of follow-up across all 727 patients, yielding a crude overall revision rate of 2.06 per 100 person-years of follow-up. The revision rate was not significantly affected by prosthesis polarity, material, or fixation technique, nor was it significantly affected by the delay of treatment. There was also no significant effect of prosthesis polarity, material, or fixation technique on postoperative range of motion. The Mayo Elbow Performance Score was only reported for half of the overall patient population, but, among those patients, the combined rate of excellent and good results was 85%. Seven percent of the overall patient population underwent secondary surgery about the elbow other than revision surgery. Twenty-three percent were reported to have 1 or more complications.
Conclusions: On the basis of our analysis of the peer-reviewed English-language literature on radial head arthroplasty from January 1940 to May 2015, there seems to be no evidence to support one type of radial head prosthesis over another. The only exception is that silicone prostheses have been shown to be biologically and biomechanically insufficient.
Level of Evidence: Therapeutic Level IV. See Instructions for Authors for a complete description of levels of evidence.
The radial head is an important secondary stabilizer of the elbow for resisting valgus stress when the primary stabilizer against valgus force, the medial collateral ligament, is injured1. This secondary stabilizing function of the radial head is particularly important following many Mason type-III fractures2,3 because of the high prevalence of associated ligamentous injuries that compromise elbow stability. The same is true following more complex elbow conditions such as so-called terrible-triad injuries, complex elbow dislocations, and longitudinal radioulnar instability. In these situations, it is imperative to replace or reconstruct, and not resect, the radial head, in order to allow healing of the damaged stabilizing soft tissues about the elbow4.
Since the introduction of radial head prostheses by Speed in 1941, various prosthetic designs have been made available5. Those designs have varied in terms of material, fixation technique, modularity, and polarity. Despite the quickly rising number of publications on radial head arthroplasty in recent years; the increasing understanding of elbow anatomy, biomechanics, and kinetics; and the evolution of surgical techniques and prosthetic designs, there is currently no evidence to support one type of radial head prosthesis over another. The only exception is that silicone prostheses have been shown to be biologically and biomechanically insufficient6.
The purpose of the present systematic review of the literature was to explore the association between prosthetic design variables and the timing of surgery and the results of modern radial head arthroplasty in order to provide evidence-based clinical recommendations. It was hypothesized that functional outcome is not different for cemented and uncemented prostheses, unipolar and bipolar prostheses, and acute and delayed treatment.
Materials and Methods
This review was intended to include patients with a minimum age of 18 years who were managed with a metallic or pyrocarbon prosthetic radial head replacement.
The primary outcome measures of interest were (1) the incidence and rate of revision, (2) postoperative ranges of motion, and (3) Mayo Elbow Performance Scores (MEPS)7. The secondary outcome measures of interest were (1) the prevalence of complications, (2) the prevalence of additional surgery about the elbow other than revision surgery, and (3) patient-reported outcome measures (PROMs), including the Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire8, Short Form-36 Health Survey (SF-36)9, American Shoulder and Elbow Surgeons questionnaire (ASES)10, Patient-Rated Elbow Evaluation (PREE)11, and Oxford Elbow Score (OES)12.
Inclusion and Exclusion Criteria
Articles written in the English language and evaluating original clinical data on primary radial head arthroplasties performed with use of metallic or pyrocarbon prostheses were considered, regardless of level of evidence. Only articles with a minimum of 10 cases and a minimum average duration of follow-up of 2 years were considered.
A study was excluded if the type of prosthesis and the duration of follow-up were not reported and were not provided by the author on request.
Search Strategy and Study Selection
A comprehensive literature search was conducted with the assistance of a clinical librarian using the following terms: radius[MeSH], radius fractures[MeSH], arthroplasty, replacement[MeSH], joint prosthesis[MeSH], radial head[tiab], replacement[tiab], arthroplasty[tiab], prosthesis implantation[tiab], and prosthesis[tiab]. The PubMed/MEDLINE and Embase databases were searched using the filters “English” and “humans” for the period from January 1940 to the date of search (May 27, 2015). The start date was chosen as the first documentation of a radial head replacement by Speed5 dated back to 1941.
Four authors independently assessed all titles and abstracts and identified eligible articles. Two authors assessed the full text of all eligible studies and made the final decision regarding inclusion. The lists of references of all eligible publications were manually checked for additional studies potentially meeting the inclusion criteria but did not yield any additional eligible studies. Disagreements were settled by discussion. With use of this strategy, 590 articles were identified (Fig. 1). After screening of title, abstract, and methodology, 43 studies were found to be potentially eligible for inclusion. The full text of those studies was analyzed, and, after application of the inclusion and exclusion criteria, 30 studies were finally included. The additional 13 articles were excluded for various reasons: 8 included only pooled data (5 for radial head arthroplasty and open reduction and internal fixation, 2 for radial head arthroplasty and radial head resection, and 1 for radial head arthroplasty, open reduction and internal fixation, and nonoperative treatment), 2 included data on cases from a larger case series that was presented more recently, 2 assessed only radiographic outcome parameters, and 1 did not specify the type of prosthesis.
Only 14 of the 30 studies included data on individual patients. The other 16 studies included only pooled data. As a consequence, analyses covering all 30 studies had to be performed on the aggregated study level, with the data on the individual patients pooled per study.
The number of patients who underwent revision surgery was expressed for each study in relation to the total number of patients in that study and in relation to the total number of person-years of follow-up per study. Revision was defined as removal of the prosthesis, or a part of it, regardless of whether a new radial head or other type of prosthesis was implanted and regardless of indication.
The various ranges of motion were expressed as an unweighted average and as a weighted average with the number of patients per study group as weights; the latter equaled the mean across the maximum of 727 patients.
For the analysis of MEPS data, studies were included only if the MEPS had been used as intended. For this analysis, only the interpreted outcomes (coded as 1 for poor, 2 for fair, 3 for good, and 4 for excellent), and not the numerical outcomes, were used and averaged per study.
Because of the substantial variability in reporting and the nature of the complications, we decided that summarizing the complications data in tabular format would provide the most information (Table I). In addition, the total number of patients with a complication was expressed for each study in relation to the total number of patients in that study. We calculated the number of complications from each individual publication and did not routinely use the number that was stated in the publication for analysis; therefore, there may be differences between the number reported in the original publication and the number reported in the current review. Radiographic observations that were considered to be complications included symptomatic loosening of the prosthesis, subluxation of the prosthesis, and periprosthetic fracture. Other radiographic findings were only considered to be complications if they required surgical intervention. As a consequence, periarticular ossification was only considered to be a complication if a resection was performed. The outcome variable of secondary surgery was analyzed in a similar fashion. Secondary surgery was defined as all surgery about the elbow, including revision surgery, performed after the index procedure.
All outcome parameters were analyzed for dependence on the prosthesis polarities, materials, and fixation techniques, both jointly and separately. In 27 of the 30 studies, only 1 combination of prosthetic polarity, material, and fixation technique (i.e., only 1 type of prosthesis) was used. In 3 studies, 2 combinations (i.e., 2 types of prostheses) were used and were considered as separate groups in the analyses. Therefore, the aggregated pooled data file used for the analyses consisted of 33 study groups rather than 30 studies. All outcome parameters were also analyzed according to delay of treatment as dichotomized in 3 alternative ways: ≤1 week versus >1 week, ≤2 weeks versus >2 weeks, and ≤1 month versus >1 month. For each study, these dichotomized delays of treatment were summarized as the proportion of the number of patients. Age (averaged per study) also was used as an explanatory variable.
The overall population of all radial head prostheses was characterized with use of descriptive statistics on the aggregated level of study group. Unweighted as well as weighted averages across study groups were calculated with use of the number of patients per study as weights. Special attention was paid to the joint outcome of prosthesis polarity, material, and fixation technique and the operational interrelationship between these modalities. Revision data, aggregated per study, were analyzed with use of generalized linear modeling. The number of revisions per study was assumed to have a negative binomial distribution, with the logarithm of the number of person-years as offset. For the effect of explanatory variables on the revision rate, a log-link function was used. Rate ratios, the dispersion parameter, and mean revision rates per category of the explanatory variables were estimated with use of the model. The dispersion parameter accounted for the extra-Poisson variability across studies. The various ranges of motion and the MEPS were analyzed with use of weighted linear regression, with the number of valid patients per study group as weights. For all analyses, a p value of ≤0.05 was considered significant. Analyses were performed with use of SPSS 21.0 (IBM) and SAS 9.2 (SAS Institute).
Thirty articles involving 727 patients were included (Table I)13-42. The number of patients per study ranged from 11 to 51. All studies were case series (Level-IV therapeutic studies). Studies originated in Europe (17 articles), North America (9 articles), Asia (2 articles), Australia (1 article), and northern Africa (1 article). The oldest article was published in 1993, and the most recent was published in 2015. Variability among the studies in terms of the reporting of patient and population characteristics was substantial. Data on prosthetic material and modularity were complete because failure to report these parameters would have resulted in exclusion of the study. The most frequently used prostheses were the Evolve Modular Radial Head (Wright Medical), the Radial Head System (Tornier), and the MoPyC radial head (Tornier), together accounting for 54% of all prostheses (Table II). Regarding implant material, 506 prostheses (70%) were made of cobalt-chromium, 107 (15%) were made of pyrocarbon, 67 (9%) were made of titanium, and 47 (6%) were made of Vitallium. Regarding polarity, 508 implants (70%) were monopolar and 219 (30%) were bipolar. Regarding fixation, 156 implants (21%) were cemented in place, 230 (32%) were press-fit, 234 (32%) were intentionally loose-fit, and 107 (15%) were fixed with an expandable stem (Table III). Only 8 of the 32 theoretically possible combinations of prosthesis polarity, material, and technique of fixation were observed (Table III). This relatively small number of observed combinations complies with the operational interrelationship between these 3 prosthesis modalities and reflects the true spectrum of available prostheses. In 3 studies, 2 combinations of prosthesis polarity, material, and fixation technique (i.e., 2 types of prostheses) were used15,34,37.
Descriptive summary statistics of the various other independent and dependent variables are presented in Table IV. For each variable, 2 means were calculated across the 33 study groups: an unweighted average and a weighted average with the number of patients per study group as weights, with the latter equaling the mean across the 727 patients. These means did not differ much from one another as no systematic relationship should exist between the number of patients and the mean of a variable per study. Sample statistics other than the mean for the total group of patients, such as median, range, and standard deviation, are not reported as they do not bear a proper relationship to underlying population parameters, with 16 studies presenting only pooled data. Age at the time of surgery was reported for 667 (92%) of 727 included patients. The weighted average age at the time of surgery was 49.8 years. Data on delay of treatment were inconsistently and variably reported. It could be determined that 262 (68%) of the 388 patients for whom the delay of treatment could be inferred from the reported data were operatively managed with a delay of ≤1 week, or that 299 (75%) of 398 patients for whom the delay could be inferred were operatively managed with a delay of ≤2 weeks, or that 424 (83%) of 508 patients for whom the delay could be inferred were operatively managed with a delay of ≤1 month. Information on delay of treatment was not reported for 46.6%, 45.3%, and 30.1% of patients with an inferred delay of ≤1 week, ≤2 weeks, and ≤1 month, respectively. The weighted average duration of follow-up was 45 months.
Primary Outcome Measures
The percentage of patients who underwent revision surgery ranged from 0% to 29% when expressed in relation to the number of patients in the studies. The number of revisions per 100 person-years of follow-up ranged from 0 to 19.83 across studies. In total, there were 56 revisions (8%) during 2,714 person-years of follow-up across all 727 patients, yielding a crude overall revision rate of 2.06 per 100 person-years of follow-up. With use of generalized linear modeling, the mean revision rate across all 33 study groups (i.e., for the entire overall population of 727 patients) based on the negative binomial distribution was 2.53 (95% confidence interval [CI], 1.53 to 4.20) per 100 person-years of follow-up. The simultaneous effect of all 8 combinations of prosthesis polarity, material, and fixation technique was not significant (p = 0.92) (Table V). For combinations with a small underlying number of patients, the imprecision of the estimates was large, as reflected by the wide CIs. Even the largest pairwise difference was not significant. Also, the separate prosthesis modalities of polarity, material, and fixation technique had no significant effect on the revision rate (p = 0.46, 0.69, and 0.98, respectively). The revision rate was also not significantly affected by delay of treatment when dichotomized at 1 week (p = 0.52), 2 weeks (p = 0.87), or 1 month (p = 0.77). It must be noted that delay of treatment was not reported in many studies. In addition, age was found to have no significant effect on the revision rate (p = 0.92).
The range-of-motion variables of flexion-extension arc and pronation-supination arc were analyzed with use of weighted linear regression (Table VI). Neither the flexion-extension arc nor the pronation-supination arc was significantly affected by the combination of prosthesis polarity, material, and fixation technique (p = 0.76 and 0.19, respectively). The flexion-extension arc and the pronation-supination arc also were not significantly affected by the separate prosthesis modalities of polarity (p = 0.88 and 0.26, respectively), material (p = 0.87 and 0.23, respectively), and fixation technique (p = 0.86 and 0.18, respectively). Delay of treatment of ≤1 week or ≤2 weeks resulted in a higher pronation-supination arc than longer delays (130.0° compared with 157.2° [difference, 27.2°; 95% CI, 13.1° to 41.2°; p = 0.001] and 129.1° compared with 154.2° [difference, 25.1°; 95% CI, 9.2° to 41.0°; p = 0.004], respectively). A delay of ≤1 month had no significant effect on the pronation-supination arc as compared with a delay of >1 month (130.8° compared with 146.3° [difference, 15.5°; 95% CI, −5.0° to 36.0°; p = 0.13]). Delay of treatment had no significant effects on the flexion-extension arc (p = 0.77, 0.61, and 0.70 for delays of ≤1 week, ≤2 weeks, and >1 month, respectively).
The average MEPS was reported for 18 of the 33 study groups (Table IV). The combined excellent and good results accounted for 85%. The standard deviation (SD) of the average MEPS across study groups was 0.26, allowing small mean differences between groups to reach significance. The combination of prosthesis polarity, material, and fixation technique had an overall significant effect on the MEPS (p = 0.038), meaning that at least 1 combination differed from the other ones (Table VII). Testing the difference of each combination with another combination (pairwise comparisons) indicated that the difference between the combinations bipolar/cobalt-chromium/cemented fixation and monopolar/pyrocarbon/expandable stem mainly contributed to this overall significant effect (p = 0.003). The p values for the separate effects of the prosthesis modalities on the MEPS were 0.13 for polarity, 0.057 for material, and 0.009 for fixation technique. Although the coded interpreted outcome for all 4 fixation techniques was on average between good and excellent, it appeared that, within this small range, press-fit fixation (mean, 3.5) and fixation with an expandable stem (mean, 3.7) resulted in better outcome scores on average than cemented fixation (mean, 3.1) (p = 0.019 and 0.002, respectively). Fixation with an expandable stem (mean, 3.7) also scored better than intentional loose-fit (mean, 3.3) (p = 0.024). It must be noted that the effect of a separate prosthesis modality is operationally confounded by effects of the other modalities on the MEPS.
Secondary Outcome Measures
There was substantial variation between studies in the reporting of complications, and the definition of complications varied among studies. A particular issue was whether or not radiographic observations were counted as complications. Therefore, it was deemed that comparing incidences and rates of complications was of no value. In the overall patient population, 182 complications were reported in 167 (23%) of the 727 patients (Table I).
The prevalence of secondary surgery, including revisions, ranged from 0% to 42% when expressed in relation to the valid number of patients in the studies. Altogether, 107 (15%) of the 727 patients underwent secondary surgery including revision surgery and 51 (7%) of the 727 patients underwent secondary surgery other than revision surgery.
PROMs were scarcely reported. Thirteen articles involved use of the DASH questionnaire; 3 involved the SF-36; 2 involved the ASES questionnaire; 2 involved the PREE; and none involved the OES.
The radial head is an important secondary stabilizer of the elbow1. Replacement of the radial head is advised in cases in which the extent of ligamentous injury calls for this secondary stabilizing function and it is not possible to reconstruct the radial head. This is the case with Mason type-III radial head fractures and more complex elbow trauma. There is currently no evidence to support one type of modern radial head prosthesis over another. The current study showed that the prevalence of revision of radial head arthroplasty, regardless of indication, ranged from 0% to 29% among the included studies. Among all patients, 8% underwent revision surgery. For the overall study population, the estimated revision rate was 2.53 per 100 person-years of follow-up. Prosthesis polarity, material, and fixation technique had no significant effect on the revision rate or the functional outcome in terms of range of motion. The combination of prosthesis polarity, material, and fixation technique appeared to have a significant effect on the MEPS; however, the MEPS was only reported for half of the patients in the overall study population. Twenty-three percent of the overall study population had some complication, and 7% underwent secondary surgery other than revision surgery.
Several limitations are recognized. Studies were included regardless of surgical indications. Because of the mixed populations in many studies and limited reporting of associated pathology, it was not possible to perform analyses stratified for indication and/or associated pathology. There was substantial variability among studies in terms of the reporting of results. This variability was particularly true for radiographic parameters. Prosthesis polarity, material, and fixation technique are not independent of each other. This resulted in only 8 possible combinations instead of the maximum of 32 possible combinations (2 different polarities, 4 different materials, and 4 different techniques of fixation) were the prosthetic design parameters to be completely independent. As stated, these 8 possible combinations reflect the true spectrum of available prostheses. Statistically, this means that prosthesis polarity, material, and fixation technique are inextricably operationally confounded with each other. Furthermore, a much larger sample size would be needed to obtain significant effects with sufficient power. Even for the largest pairwise contrast in revision rates between combinations of prosthesis polarity, material, and fixation technique, a sample size at least 3 times larger would be needed.
In the overall population of the current study, there were 56 revisions (8%) among the 727 patients, including those with both monopolar (n = 44) and bipolar (n = 12) prostheses. Thirteen (23%) were for loosening; 11 (20%), for overstuffing; 10 (18%), for subluxation; 8 (14%), for stiffness; 7 (13%), for lateral elbow pain; 3 (5%), for dissociation of the prosthesis; 2 (4%), for infection; 1 (2%), for malposition; and 1 (2%), for fracture of the stem of the prosthesis. One revision for loosening involved a cemented prosthesis, and 12 involved a press-fit prosthesis13,24,29,34. Osteolysis around the stem has been described in about 50% of all patients with a press-fit radial head prosthesis15,19,23,24,26,29,31,34,37, with a prevalence of 17% to 100% in the various studies included in this review. In most cases, this osteolysis was deemed asymptomatic. It is noteworthy that 9 of the 12 failures of press-fit prostheses occurred in a single case series by Flinkkilä et al. and were due to lateral elbow pain and implant loosening24. Although the authors attributed this finding to poor osseous ongrowth, it is not clear why other studies involving the same or similar prostheses demonstrated substantially lower failure numbers. Biomechanical studies have suggested that monopolar prostheses better restore elbow stability than bipolar prostheses and it was therefore suggested that monopolar prostheses would better resist radiocapitellar subluxation43,44. However, Berschback et al. and Rotini et al. could not confirm this hypothesis in comparative clinical studies15,34. In the overall population of the current study, 5 revisions for subluxation involved a monopolar prosthesis and 5 involved a bipolar prosthesis. When replacing the radial head, restoration of the length of the radius is important. Overstuffing may result in increased radiocapitellar pressure, which may in turn lead to attrition of the capitellar articular cartilage and pain45. Conversely, failure to restore the length of the radius may result in instability. It has been proposed that the proximal edge of the lesser sigmoid notch of the ulna with the forearm in neutral rotation can be used as a reference for the articulating surface of the prosthesis46. It has been postulated that restoring radial length may be easier to accomplish with a modular prosthesis. This hypothesis could not be confirmed in the current study; of the 11 revisions that were performed because of overstuffing in the overall population of the current study, 10 involved modular prostheses and 1 involved a monoblock prosthesis.
There was large variation among studies in terms of the quality and quantity of the reported data. Not infrequently, important methodological information or important outcome parameters were lacking. This heterogeneity in data hampers sound comparison of studies and renders it impossible to conduct a formal meta-analysis. This problem also has been recognized for total elbow arthroplasty47. It is likely that the peer-review process is falling short in ensuring uniformity and quality in data reporting. A discussion about the institution of guidelines for standardized reportage of clinical outcomes seems appropriate. This also holds true for radiographic parameters. Radiographic findings were reported with too great a variability to allow for a structured analysis, and for that reason no radiographic outcome parameter was used in this study.
It has been recognized that associated injury about the elbow, forearm, or wrist is very likely to affect the outcome after radial head arthroplasty. However, in the current study, it became apparent that associated injuries were so scarcely and variably reported that no reliable analysis could realistically be attempted. It is assumed that such associated injuries are often missed during physical examination. Magnetic resonance imaging (MRI) studies have shown that 76% to 96% of radial head fractures may be associated with other injuries about the elbow45,48-50. Also, if adequate reconstruction or replacement of the radial head is performed in a timely fashion, many of these injuries may heal without clinical consequences and may not be diagnosed at all.
In summary, on the basis of our analysis of the peer-reviewed English-language literature on radial head arthroplasty from 1940 to 2015, there seems to be no evidence to support one type of radial head prosthesis over another, with the exception of silicone prostheses, which have been shown to be biologically and biomechanically insufficient.
Investigation performed at the Department of Orthopaedic Surgery, Amphia Hospital, Breda, the Netherlands
Disclosure: No funding was received related to this research from any party in the public, commercial, or non-profit sector. The Disclosure of Potential Conflicts of Interest forms are provided with the online version of the article.
- Copyright © 2016 by The Journal of Bone and Joint Surgery, Incorporated