Patient- and Procedure-Specific Variables Associated With Removal or Revision of Radial Head Arthroplasty

Abstract

Introduction

The purpose of this retrospective review was to identify risk factors associated with removal or revision following radial head arthroplasty.

Methods

Patients undergoing primary radial head arthroplasty between 2009 and 2015 with a minimum follow-up of 1 year were identified. Descriptive and bivariate statistics were used to analyze the characteristics of patients requiring implant removal or revision and multivariable analysis was performed to calculate hazard ratios.

Results

There were 312 patients included in the final cohort with a median follow-up of 3.8 years. Thirty-five patients (11.2%) underwent prosthesis removal or revision. There was an increased percentage of implants removed or revised in patients under age 40, with surgery performed for chronic indications compared to acute trauma, and with the use of press-fit stems compared to polished.

Discussion

It appears younger patient age, chronic surgical indications, and certain aspects of prosthesis design may influence rates of removal or revision.

Introduction

Radial head fractures are common injuries, accounting for approximately 30% of all elbow fractures. ¹ In complex elbow trauma involving radial head fracture and associated ligamentous injury, restoration of proximal radial anatomy is paramount to restoring elbow stability. ² In such cases, radial head fractures can be addressed by open reduction and internal fixation (ORIF) or replacement of the fragmented pieces with a radial head arthroplasty. Outcomes of radial head ORIF versus arthroplasty have been investigated and several studies have demonstrated superior results with arthroplasty, particularly in cases with > 3 fracture fragments. ^3-5 Radial head arthroplasty may also be used as a reconstructive option to treat radial head or neck nonunion, malunion, and posttraumatic arthritis. ^6,7

Although the majority of patients do well following radial head arthroplasty, some studies have reported revision rates approaching 30%. ⁸ The most common reasons for implant removal are pain and limitation of motion. ^8,9 Known complications of radial head arthroplasty include elbow stiffness, loosening, implant disassembly, capitellar erosion, and deep infection. ^10,11 Technical factors with regard to implant sizing are well-recognized reasons for failure, but otherwise the current evidence is contradictory as to what other factors may be associated with implant failure. ¹²

The current body of literature investigating radial head arthroplasty is insufficient to guide surgeons on the potential for implant removal because further information is necessary to define patients who are more likely to have complications. The purpose of this study was to determine if there are patient or procedure factors that are predictive of radial head removal.

Methods

This was a retrospective cohort study conducted within a single region of a large integrated health care system, the membership of which is demographically similar to the underlying population. ^13,14 Patients undergoing primary radial head arthroplasty between 2009 and 2015 were identified using an EPIC operating room database with an internal coding system (EPIC Systems Corp, Verona, WI, USA). Physician-led electronic chart review confirmed inclusion eligibility for each identified potential patient.

The study period spanned from January 1, 2009 to December 31, 2017. Patients under the age of 18 or who left the health plan < 1 year following date of index surgery were excluded. Radial head arthroplasties done as part of radiocapitellar or total elbow arthroplasty were also excluded, as were cases utilizing cement for fixation.

Baseline demographic and clinical characteristics were obtained from the electronic medical record, including age at time of index surgery, sex, body mass index (BMI; computed from height and weight measurements obtained closest to the index date), and Charlson Comorbidity Index ¹⁵ (using a 1-year pre-index date capture of diagnosis and procedure codes linked to inpatient and outpatient encounters). Diabetes status at the index date was identified using an institutional diabetes registry.

The reason for surgery was identified through chart review of the electronic medical recordby 2 of the authors (both fellowship-trained orthopedic surgeons) using a structured data collection tool. Indications were categorized as treatment for acute trauma or a chronic condition. Chronic conditions were defined as primary radiocapitellar arthritis, failed prior ORIF, or sequela of trauma > 6 weeks from injury. Additional procedures performed at the time of primary radial head arthroplasty were identified, including lateral ulnar collateral ligament (LUCL) repair or reconstruction, olecranon fracture ORIF, and coronoid fracture ORIF. The design of the radial head prosthesis was recorded and categorized as smooth polished implants (designed to have some rotational motion within the proximal radius) versus press-fit implants (designed to stay rigidly fixed within the proximal radius), and bipolar versus unipolar.

The primary outcome of this study was removal or revision of a radial head implant. Secondary outcomes were time from index surgery to removal or revision, primary reason for implant removal or revision (pain, stiffness, infection, instability, and other), and additional surgical procedures performed at time of implant removal or revision.

Bivariate statistics were used to determine which factors were crudely associated with implant removal or revision. Categorical variables were reported using frequencies and proportions. Associations with implant removal or revision were tested with χ² tests. Nonnormally distributed continuous variables (age and BMI) were reported using median and interquartile range statistics, and comparisons were made using Kruskal-Wallis tests. A Kaplan-Meier curve was constructed to present removal or revision rates after initial radial head arthroplasty. Follow-up for time-to-event analysis was defined as the time from the date of the index procedure to the date of implant removal or revision, membership termination, death, or study end date, whichever came first. In addition, bivariate and multivariable Cox regression analyses were performed to determine which demographic and clinical variables were independently predictive of removal or revision of implants. Predictors used in the model included those that were statistically significant in bivariate analysis as well as those generally accepted in the literature to be potential cofounders. Final multivariable Cox proportional hazards regression models were constructed adjusting for age category (< 40 years, ≥ 40 years), sex, type of implant, surgical indication, and additional procedures performed (LUCL repair or reconstruction, ORIF olecranon). Hazard ratios (HRs) with Wald 95% confidence intervals (CIs) and corresponding p values were reported. Additional surgical procedures performed at the time of implant removal and reason for implant removal were reported using descriptive proportions, as this sample was expected to be small. P values < 0.05 were considered significant.

Institutional Review Board approval was obtained prior to the inception of this study.

Results

The study investigators reviewed 432 potential radial head arthroplasty cases. Eighty-four surgical procedures did not meet the case definition. Of the 348 remaining patients, 36 were excluded for not meeting the membership requirement (N = 34) and/or for being cemented procedures (N = 3). The final study sample comprised 312 patients, with procedures performed by 100 surgeons (range: 1–19 procedures) at 24 medical centers. Median follow-up was 3.8 years. Prosthesis survival was 90% at 2 years, with 35 patients (11.2%) having either removal or revision of a radial head prosthesis during the study period (Figure 1). Median time from implantation to removal was 8.8 months, with 21 of 35 patients having their implants removed or revised within 1 year of surgery.

Figure 1:

Kaplan-Meier survival for revision or removal following radial head arthroplasty. The shaded area represents the 95% confidence limit.

The median age of the cohort was 58 years (interquartile range: 45–66). There were 185 women (59.3%). Patients under the age of 40 had a higher proportion of implants undergoing removal or revision when compared with patients aged 40 years or higher (22.6% vs 8.9%, p < 0.01). Other patient-specific variables did not appear to be associated with removal or revision with no more than a 5.2% difference in revision proportions across categories (Table 1). Patients undergoing surgery for a chronic condition were more apt to undergo removal or revision as opposed to those who had surgery for acute fractures (25.9% vs 9.8%, p = 0.01). Press-fit stems were also associated with an increased proportion of removal or revision (12.7% vs 2.2%, p = 0.04). Polarity of the prosthesis and associated procedures done at the time of primary surgery did not show any significant association with removal or revision with the numbers available (Table 2).

Table 1:

Patient-specific risk factors for removal or revision

Risk factors	Prosthesis removal or revision			Percent removed or revised	p value a
	Overall N = 312 (%)	No n = 277 (%)	Yes n = 35 (%)	11.2%
Age, median (IQR)	58 (45–66)	58 (46–66)	51 (32–63)		0.09
Categorical age					< 0.01
< 40 years	53 (17.0)	41	12	22.6
≥ 40 years	259 (83.0)	236	23	8.9
BMI, median (IQR)b	29.0 (25.7–34.2)	29.2 (25.5–34.2)	28.2 (25.7–34.7)		0.65
Categorical BMI					0.56
< 25	67 (21.9)	60	7	10.4
25–29	97 (31.7)	83	14	14.4
30–34	78 (25.5)	72	6	7.7
≥ 35	64 (20.9)	56	8	12.5
Sex, n (%)					0.93
Men	127 (40.7)	113	14	11.0
Women	185 (59.3)	164	21	11.4
Diabetes, n (%)					0.55
No	281 (90.1)	248	33	11.7
Yes	31 (9.9)	29	2	6.5
Charlson Comorbidity, no. (%)					0.97
0	220 (70.5)	195	25	11.4
1–3	81 (26.0)	72	9	11.1
≥ 4	11 (3.5)	10	1	9.1

^a Age and BMI p values calculated using Kruskal-Wallis tests. Categorical comparisons made with χ² tests. Fisher’s exact test used for diabetes variable.

^b Missing BMI values for 6 patients.

BMI, body mass index; IQR, interquartile range.

Table 2:

Procedure-specific risk factors for removal or revision

Risk factors	Prosthesis removal or revision			Percent removed or revised	p value a
	Overall N = 312 (%)	No n = 279	Yes n = 35
Diagnosis, no. (%)					0.01
Acute fracture	285 (91.3)	257	28	9.8
Chronic condition	27 (8.7)	20	7	25.9
Polarity, no. (%)					0.99
Unipolar	289 (92.6)	256	33	11.4
Bipolar	23 (7.4)	21	2	8.7
Fixation, no. (%)					0.04
Polished	45 (14.4)	44	1	2.2
Press-fit	267 (85.6)	233	34	12.7
Additional Procedures b , no. (%)
LUCL repair or reconstruction	102 (32.7)	87	15	14.7	0.17
ORIF coronoid	51 (16.3)	46	5	9.8	0.73
ORIF olecranon	65 (20.8)	61	4	6.2	0.14

^a p values calculated using χ² tests. Polarity p value calculated using Fisher’s exact test.

^b More than one additional procedure may have been performed at time of surgery.

LUCL, lateral ulnar collateral ligament; ORIF, open reduction and internal fixation.

In a multivariable Cox proportional hazards regression, patient age of less than 40 years (adjusted HR, aHR = 2.47, 95% CI = 1.17–5.25) and chronic surgical indication (aHR = 2.67, 95% CI = 1.10–6.49) were predictive of increased removal or revision compared with older patients and acute surgical indication, respectively (Table 3).

Table 3:

Multivariable Cox proportional Hazard model of risk factors

Risk factors	Adjusted hazard ratio	95% confidence limits	p value
Age < 40 years	2.47	1.17–5.25	0.02
Female sex	1.30	0.64–2.67	0.47
Chronic condition	2.67	1.10–6.49	0.03
Press-fit prosthesis	5.13	0.70–37.93	0.11
Additional procedure
ORIF olecranon	1.73	0.86–3.51	0.13
LUCL repair or reconstruction	0.62	0.21–1.82	0.38

LUCL, lateral ulnar collateral ligament; ORIF, open reduction and internal fixation.

Two press-fit radial head prostheses implanted during the study period have since been voluntarily recalled. The DePuy Synthes Radial Head Prosthesis System (Synthes USA, West Chester, PA) has shown high rates of postoperative loosening ¹⁶ and was recalled in January 2017. Likewise, poor results have also been reported with the now-recalled Stryker rHead Radial Head and Uni-Elbow System (Stryker, Kalamazoo, MI). ¹⁷ During the study period, 9/46 Synthes and 6/20 Stryker radial head implants were removed or revised. Compared to other press-fit stems (N = 201), use of Synthes implants (aHR = 2.5, 95% CI = 1.1–5.6) and Stryker implants (aHR = 3.6, 95% CI = 1.4–9.0) was associated with increased removal or revision, in a model adjusting for age and indication (chronic vs acute injury). It is certainly possible that these 2 implants could negatively impact the overall survival of press-fit prostheses in our study. As such, we further investigated press-fit implants with these 2 prostheses excluded. We noted a 9.5% removal or revision proportion for the remaining press-fit prostheses compared to 2.2% in the polished stems, which did not reach statistical significance with the numbers available (Fisher’s exact p = 0.14).

The overwhelming majority of patients had removal or revision of their implant for pain (23/35) and/or stiffness (17/35) (Table 4). Table 5 lists the additional procedures done at removal or revision surgery, with LUCL repair or reconstruction (6/35) being the most common.

Table 4:

Reasons for removal or revision

Indication a	No. of total, N = 35
Pain	23
Stiffness	17
Instability	7
Excision of heterotopic ossification	3
Infection	2
Periprosthetic fracture	1

^a More than 1 indication for removal or revision may be present.

Table 5:

Additional procedures done at time of removal or revision

Procedure a	No. of total, N = 35
LUCL repair or reconstruction	6
Capsular release	4
Other hardware removal	3
Excision of heterotopic ossification	3
Ulnar nerve decompression or transposition	3
Placement of external fixator	3
Irrigation and debridement for infection	2
Medial ligament repair or reconstruction	1

^aMore than 1 additional procedure may have been performed at time of removal or revision.

LUCL, lateral ulnar collateral ligament.

Discussion

Our study demonstrated an 11.2% revision or removal proportion for radial head implants at a median follow-up of 4.2 years. The overwhelming majority of our patients who elected to have implant removal or revision did so for complaints of stiffness or pain, similar to other published findings. ^8,9,18 However, it deserves to be noted that it is difficult to determine if complaints of stiffness are secondary to the presence of a radial head implant or are simply due to the initial injury. At the time of revision or removal, we found the most commonly performed additional surgery was LUCL repair or reconstruction (17.1%), although whether this is due to ligament damage incurred during implant removal is unknown.

Prior studies investigating patient-specific factors for failure have yielded inconsistent findings. Several studies have not found an association between characteristics such as age and sex with radial head failure. ^19-21 Duckworth et al, ⁸ however, analyzed the outcomes of 105 patients with uncemented monopolar prostheses and noted younger age was a risk factor for implant removal or revision. We also found that younger age was a risk factor for prosthesis removal or revision, which may be due to the higher demands these patients place on their elbow. As a result, surgeons should carefully counsel their younger patients prior to placing a radial head implant and alert them to the fact they may require additional surgery in the future.

Radial head replacement is most often used for acute injuries, and it is also an option for posttraumatic and other chronic conditions. ^6,7,12,22-24 In one of the few studies comparing radial head arthroplasty for acute and chronic injuries, Chapman et al ²³ reviewed a series of 16 patients and reported greater subjective disability and worse motion in those operated on for chronic conditions, although no patients required radial head removal or revision. In our study of 312 patients, we noted that radial head replacements performed for indications other than acute fracture had an increased proportion of removal or revision. Our findings seem to suggest that radial head prostheses placed for chronic indications may be predisposed to removal or revision, and as such these patients may be better served by a different procedure.

Studies examining the utilization of bipolar radial head prostheses have reported inconsistent results. ^{17,18,22,24-26} In their meta-analysis of over 1000 implants, Kachooei et al ¹⁸ noted a 10% removal or revision proportion for monopolar prostheses compared to 3.7% for those of a bipolar design, although they acknowledge this may be surgeon-dependent as bipolar implants can be more difficult to remove. We did not note a substantial difference in removal or revision rates when comparing bipolar with monopolar implants. It is possible that this may be due to sample size, as only 23 of the 312 radial head arthroplasties implanted in our study were of a bipolar design.

Although smooth-stemmed prostheses have demonstrated fairly reliable results, ²⁷ the findings of studies examining the outcomes of press-fit stems have been less clear, ^16,17,28-33 with some authors suggesting differences within various designs of press-fit stems. ³² One recent study reported 24% of press-fit prostheses were removed with a mean time to loosening of 11 months. ²⁹ Similarly, Lobo-Escolar et al ³³ noted a 26.7% reoperation rate with press-fit implants, with a prosthesis survival rate of 69.5% at 24 months. Yet, prior studies directly comparing polished versus press-fit stems have failed to show substantial differences regarding implant removal or revision. ^21,34,35 When comparing press-fit with polished stems, we found that press-fit stems underwent removal or revision at a significantly higher proportion (12.7% vs 2.2%, p = 0.04), and there was only removal or revision of 1 polished prosthesis. However, when we excluded the recently recalled prostheses, there was only a trend toward an increased proportion of removal or revision for press-fit stems (9.4% vs 2.2%, p = 0.14). Although potentially clinically significant, this finding did not reach statistical significance, which could be due to the low overall event rate. It remains unclear if press-fit implants as a whole are problematic, but we agree with the findings of Vannabouathong et al ³² and believe that it appears the specifics of press-fit design may impact outcomes following radial head arthroplasty.

Radial head fractures often do not occur in isolation and can be part of a more complex injury pattern. Some have suggested that press-fit implants may perform less reliably in more complicated fracture patterns, such as terrible triad and Essex-Lopresti injuries. ³² However, Strelzow et al ³⁶ reviewed 148 patients who underwent radial head arthroplasty for either “simple” or “complex” injury patterns, and the authors did not notice any differences in objective findings or patient-reported outcome measures between the 2 groups with a minimum 1-year follow-up. Lott et al ³⁷ noted a 10º loss of supination in those patients undergoing radial head arthroplasty as part of an unstable fracture pattern, but the authors did not see any difference in complications or implant survival between the 2 groups. In contrast, Nolte et al ²¹ reported that use of an external fixator was associated with revision surgery when utilizing a radial head arthroplasty. In our current study, we noted that neither LUCL repair or reconstruction, ORIF of the olecranon, nor ORIF of the coronoid were associated with an increased chance of prosthesis removal or revision.

This study of over 300 patients is one of the largest to date investigating factors associated with removal or revision of radial head implants. The inclusion of multiple prosthesis types allowed us to investigate how implant design may influence removal or revision rates. Furthermore, our study included many surgeons from several hospitals, thus increasing the generalizability of our findings and potentially minimizing surgeon or hospital preference as a factor in the decision for prosthesis removal or revision.

This study has several weaknesses that must be considered. This study is retrospective in nature and thus may be subject to the introduction of bias and precluded our ability to perform a power analysis. Additionally, it is certainly possible that patients may have had further surgery outside of the health plan after terminating membership, although health plan attrition is notably low. Our 1-year minimum follow-up, however, should minimize this impact as radial head prosthesis removal or revision tends to occur within the first year of surgery, according to findings of both our current study and previous publications. ^8,19 It is possible, however, that with a longer follow-up we would have captured additional revisions or removals for our analysis. We also did not take surgeon experience or practice patterns into account, which certainly may influence prosthesis removal or revision rates. ¹⁹ We also acknowledge that our study design did not include clinical outcome measures such as range of motion, patient-reported outcome scores, or postoperative radiographs. Although radiograph review may have been useful in identifying technical errors such as overstuffing, some authors have advised caution in using postoperative radiographs to make treatment decisions regarding radial head arthroplasty due to poor interobserver reliability. ³⁸

In summary, patient- and procedure-specific factors may play more of a role than previously thought regarding revision of radial head arthroplasty. Younger patient age and chronic surgical indications seem to be associated with lower survival rates of radial head arthroplasty. We identified 2 underperforming press-fit implants that have since been recalled, although further studies are needed to determine the role that implant design may play in implant survival. Surgeons should keep these risk factors in mind when deciding to perform radial head arthroplasty and counsel their patients accordingly.