Removal of implants after open reduction and internal fixation (ORIF) of tibial plateau fractures has been reported, with rates ranging from 6% to more than 40%.1–4 Although implant removal is commonly performed to alleviate localized postoperative pain,2,3,5 the efficacy of pain reduction has been mixed.6–9 A reduction in these procedures would attenuate the burden of additional surgery on patients,10 but research investigating modifiable factors associated with symptomatic implant removal has been sparse.11 To the current authors' knowledge, no study has investigated this question regarding tibial plateau fractures.
The authors' objective was to determine the index treatment factors associated with removal of implants after ORIF of tibial plateau fractures. Several radiographic, demographic, and injury measures were explored. On the basis of previous findings3,4,11 and clinical experience, the authors hypothesized that greater protrusion of implants, closer proximity of fixation to the joint, large implant size, and injury severity are associated with symptomatic implant removal.
Materials and Methods
This retrospective case-control study received institutional review board approval. Patients with tibial plateau fractures who underwent ORIF at the authors' level I trauma center from 2007 to 2016 were identified from a prospectively collected hospital billing database. Indication for implant removal was localized pain over the implant. Removals for other reasons, such as infection, nonunion, or staged total knee arthroplasty (TKA), were excluded, leaving 82 cases of symptomatic implant removal (Figure 1). From the remaining population of patients with tibial plateau fractures, cumulative sampling was used to select control patients based on a 1:2 case-to-control ratio. Several of the control patients were excluded because of ipsilateral intramedullary nail, fixation without a plate, age younger than 18 years, missing radiographs, planned TKA, or death during hospitalization, leaving a control group of 132 fractures. The mean age of the 214 study patients was 46 years (SD, 15), the majority of the patients were male (n=126; 59%) and white (n=139; 65%), the mean Injury Severity Score (ISS) was 12.5 (SD, 7.5), and the mean body mass index (BMI) was 29 kg/m2 (SD, 9).
CONSORT diagram. Abbreviations: ORIF, open reduction and internal fixation; TKA, total knee arthroplasty.
The primary outcome was symptomatic implant removal. Symptomatic was defined as localized pain over the implant with an otherwise unknown cause. Removal of implants was confirmed by viewing post-removal radiographs. Localized discomfort as the cause was verified in the preoperative clinic note.
Demographics and injury characteristics were selected based on previous literature3,4,11 and clinical relevance. Radiographic measurements were based on 3 categories: proximity of implant to joint, amount of implant protrusion, and implant size. Proximity of implant to joint was measured by the closest plate or screw to the joint surface on the anteroposterior view (Figure 2). Amount of implant protrusion was quantified by counting the number of screws with tips protruding beyond the far cortex and measuring the amount of protrusion of the most protruding screws on the anteroposterior view (Figure 2). Implant size was measured by the lengths of the plates on the lateral view. Three authors (C.C.S., D.C., M.B.) measured these parameters from radiographs using PACS IMPAX 6.5 Client software (Agfa-Gevaert NV, Mortsel, Belgium). To minimize measurement error, the authors selected radiographs that provided the best aligned view of the joint surface and implant, considering the variability in immediate postoperative radiographic views. Demographics and injury characteristics were abstracted from the electronic medical records.
Measurements on anteroposterior (A) and lateral (B) radiographs.
Patient demographics, injury characteristics, and implant characteristics were summarized as means with 95% confidence intervals (CI) if continuous and as counts with proportions if categorical. Comparison based on primary outcome was performed, and unadjusted P values were obtained using Student t tests or chi-square tests as appropriate (Table 1).
Patient Demographics and Injury and Implant Characteristics
Current recommendations for prognostic models suggest a minimum of 5 to 15 cases per prognostic variable.12,13 The 82 cases included in the current study provided adequate statistical power to test approximately 5 to 16 variables in the authors' full model. Previous literature has identified associations between age, open fractures, and bicondylar fractures with removal of implants in cases of tibial plateau fractures.3,4 Proximity to the joint, screw protrusion, and sex have been identified as strong predictors in cases of retrograde femoral nail insertion.11 Body mass index and ISS were associated with removal of implants in cases of tibial shaft fractures treated with intramedullary nails.14 Number of protruding screws resulted in a higher area under the curve (AUC) than screw protrusion and was included. The authors included the above variables and total plate length in a multivariable logistic regression model.
Backward stepwise elimination based on a minimum Akaike information criterion with whole effects was used to select covariates for inclusion in the reduced prognostic model. The association between the included covariates and the removal of tibial plateau fracture implants was determined using a multivariable logistic regression model and was reported using adjusted odds ratios (OR) with 95% CI. The discriminatory capability of the model was calculated using the AUC statistic. Data were missing for fewer than 5% of BMI observations and were imputed using multiple imputations.15 No other data were missing in the prognostic model. All statistical analysis was performed using JMP Pro version 13 software (SAS Institute Inc, Cary, North Carolina) and Stata version 14.2 software (StataCorp LLC, College Station, Texas).
Implants were removed because of pain in 82 (9%) of the cases (Figure 1). Implant removal most commonly involved the entire implant (50 of 82 cases) or the plate with several screws (21 of 82 cases) rather than screws alone (11 of 82 cases). The authors' unadjusted analysis showed no differences in demographics between the removal group and the control group (Table 1). Mean follow-up was longer for the removal group than for the control group (23.8 vs 9.5 months, P<.01). Mean time to removal was 18.3 months (SD, 13.1). The removal group had more bicondylar fractures (73% vs 57%, P=.02) but similar mean ISS (13.2 vs 12.0, P=.25), proportion of contralateral fractures (26% vs 22%, P=.54), and proportion of open fractures (5% vs 9%, P=.26) compared with controls. Significant radiographic differences were observed in the number of protruding screws per fracture (3.5 vs 2.4, P≤.01) and the maximum distance of screw protrusion (3.9 mm vs 3.1 mm, P=.01) between the two groups.
The OR and coefficients for the reduced prognostic model are included in Table 2. Each additional protruding screw increased the likelihood of symptomatic implant removal by 32% (OR, 1.32; P<.001). Bicondylar fractures were 113% more likely to require symptomatic implant removals (OR, 2.13; P=.02). Conversely, increased BMI was associated with a decrease in the likelihood of symptomatic implant removal (OR, 0.84 per 5 kg/m2; P=.05). Associations that approached statistical significance were observed with decreased age (OR, 1.22 per 10 years; P=.06) and closed fractures (OR, 2.97; P=.09).
Multivariable Prognostic Model for Implant Removal After Tibial Plateau Fixation (n=214)
The prognostic model had a discriminative ability of AUC 0.71 (95% CI, 0.63–0.77) (Figure 3). Based on AUC interpretation guidelines suggested by Kleinbaum and Klein,16 the model had fair predictive value. The goodness of fit was R2=0.10.
Receiver operating characteristic curves based on a full and reduced model (area under the curve=0.71).
To the authors' knowledge, this is the first study to describe radiographic predictors of symptomatic implant removal after ORIF of tibial plateau fractures. The authors observed that symptomatic implant removal was associated with each additional protruding screw, bicondylar fractures, and lower BMI. Nominal increases in the likelihood of symptomatic implant removal were observed in younger patients and patients with closed fractures. The compounding likelihood of symptomatic implant removal with each additional protruding screw in the construct highlights the importance of selecting optimal implants for fixation of tibial plateau fractures.
Previous literature on radiographic predictors of implant removal after fracture is limited.11,14,17 The current authors' hypothesis that implants closer to the joint and with more protrusion are associated with higher removal rates was driven by clinical expectations and previous literature on retrograde femoral nails.11 Although their results showed a strong signal with protruding screws, distance to the joint hinted at this trend in bivariate analysis (3.9 mm vs 4.6 mm, P=.22) but was not significant in multivariable analysis. Previous investigation into removal of implants after ORIF of tibial plateau fractures has focused exclusively on demographics and injury characteristics.3,4 Those studies found an association between bicondylar fractures and an increased odds of implant removal, consistent with the current authors' results (OR, 2.1; P=.02). Henry et al4 found that decreased age was associated with increased odds of removal, also in agreement with the current authors' prognostic multivariable model (OR, 1.22 per 10 years; P=.06). Results from tibial shaft fractures treated with intramedullary nails14 showed an association between lower BMI and implant removal, which is supported by the current authors' findings (OR, 1.19 per 5 kg/m2; P=.05).
The finding that each additional screw protruding from the far cortex increased the odds of removal by 32% emphasizes that consideration should be given to placing screws without excess length. In contrast, screws that do not sufficiently engage the far cortex can lead to inadequate fixation,18,19 with loss of reduction and poor clinical outcomes.20,21 Guidance in the literature for radiographic engagement in the far cortex and optimal screw length is sparse. A study of retrograde femoral nail insertion concluded that excess length of proximal interlocking screws would not come at a cost to patients,22 although the authors based that on the assumption of no symptomatic removals. Further, they conceded that the clinical relevance of screws that are too short is unknown. In contrast, Hamaker et al11 found a significant relationship between longer length of the most distal screw and removal rates if the screws were within 40 mm of the joint. A study of patients with tibial plateau fractures reported a fixation failure rate greater than 30%, but the authors did not investigate whether implant characteristics and screw length affected that outcome.23 The current authors were unable to find any literature on tibial plateau fractures that investigated the relationship between screw length and clinical outcomes.
Although the authors' unadjusted analysis showed a significant relationship between screw protrusion and removal, the difference of less than 1 mm between groups is unlikely to be clinically significant. However, the association between removal and additional protruding screws was significant in adjusted analysis and is a readily modifiable finding. The authors attempted to find a number of screws above which all patients had their implants removed and a number below which no patients had their implants removed, but their data provided no meaningful cutoffs. The same question was asked regarding amount of screw protrusion, which also failed to provide minimum or maximum values. Regardless, these results suggest a potential clinical benefit to selecting screws that do not extend beyond the far cortex.24 Further studies should investigate radiographic evidence of engagement with the far cortex to guide minimum screw length. These minor adjustments have the potential to attenuate the likelihood of removal operations and the subsequent risk of complications and further time off work10 for patients with tibial plateau fractures.
The current study supports previously presented evidence that bicondylar fractures are associated with an increased likelihood of implant removal.3,4 One study4 concluded that higher energy fractures were the underlying cause. The current study's data are not in complete agreement, considering that ISS was not significantly different between the removal and non-removal groups. A supporting possibility is the increased association with medial plates in bicondylar fractures. In the current study population, bicondylar fractures were more frequently treated with medial plates (38% vs 22%, P=.01). However, medial plates were not significantly associated with increased rate of removal (P=.37), so this is unlikely to fully explain the association. A second possibility is that the additional complexity of reduction and likely greater articular involvement is more irritating to patients, increasing their willingness to undergo an additional operation that might provide relief.
The association between lower BMI14 and increased removal and the trend toward decreased age4 (P=.06) can be explained by a reduction in mobility in the older and higher BMI patient populations. Further, higher BMI is associated with a thicker layer of adipose tissue surrounding the joint, and the increased superficial soft tissue might reduce irritation from superficial rubbing of the implant. Older and heavier patients might also have an increased surgical risk assessment, reducing the likelihood of reoperations as decided by the care team. Regardless of the underlying cause, these confirmatory data regarding non-modifiable associations add confidence to previously noted associations.
This study had several strengths. These are the first data to address the effect of implant characteristics on the rate of implant removal in cases of tibial plateau fractures. The authors' measurements are easily reproducible using radiological software available at many medical centers, and the large sample provided power to detect several predictors of removal. The authors radiographically confirmed each removal procedure and its noted indication for pain, excluding the possibility of overestimation of removal rates. This study also had large heterogeneity of surgeons (n=12) and implant manufacturers (DePuy Synthes, West Chester, Pennsylvania; Stryker Corporation, Kalamazoo, Michigan; Smith & Nephew, Memphis, Tennessee; and Zimmer Biomet, Warsaw, Indiana), reducing sources of bias.
However, the limitations of this study's design should be considered. It is possible that some patients in the control group elected to have their implants removed at another center, leading to a potential misclassification for some of the control group outcomes. The authors assume the misclassification to be non-differential to the prognostic variable, and, therefore, the imperfect sensitivity does not bias the OR. The authors' center typically conducts follow-up of patients who have tibial plateau fractures for 12 months or until they have shown uncomplicated recovery. If patients expressed concerns about pain and considered implant removal before that time, it is likely that follow-up would have continued; however, this is not always the case. Further, implant removal surgery is not typically offered for several months after initial fixation, explaining the longer follow-up in the removal group. In contrast, the random sample of controls is representative of uncomplicated recovery after ORIF, limiting the need for further follow-up in that group. Additionally, the radiographic parameters were measured using standard imaging software (Figure 2). Although this simplistic technique increases the replicability of the study, it reduces measurement precision. Further, the reviewers were not blinded to outcomes, and although efforts were made to ensure consistent measurement across cases and controls, this could be a source of bias.25
A further consideration in interpreting these results is surgeon bias. If it is already clinically believed that protruding screws cause pain,19,26 this could partially explain higher removal rates in those with greater radiographic protrusion. Finally, this study did not follow patients after removal of implants to assess whether symptoms improved. This is a distinct question that has not been definitively answered in the literature. Several studies have addressed outcomes after removal of implants from other fractures, but the results have been mixed.6,7 To the current authors' knowledge, only a single study has reviewed outcomes after removal of implants from tibial plateau fractures.8 Garner et al8 did not show a significant difference in pain after removal; however, that study might have lacked the power to detect a difference (n=75). The efficacy of this procedure for reducing pain in cases of tibial plateau fractures has yet to be definitively established, highlighting the importance of efforts to identify modifiable predictors of higher removal rates. A study that prospectively acquires both risk factors and pain outcomes after removal of implants from tibial plateau fractures would be an important step toward that end.
Hospitals throughout North America perform implant removal procedures at varying rates. Prospective studies have reported rates ranging from 6%3 to more than 40%.1 The largest study in the literature reported a 2-year rate of 18% using Ontario billing databases.4 In a time of increasing focus on cost-effective patient care,27,28 a reduction in removal procedures represents an attainable reduction in health care spending. Further, the lack of sufficient evidence for the efficacy of pain reduction supports efforts to avoid unnecessary removal procedures. The current study's results suggest that minimizing the number of screws that protrude from the far cortex without compromising fixation might reduce the burden of subsequent surgery in this fracture population. Further studies should confirm these predictors to guide optimal fixation.
- van Dreumel RL, van Wunnik BP, Janssen L, Simons PC, Janzing HM. Mid- to long-term functional outcome after open reduction and internal fixation of tibial plateau fractures. Injury. 2015;46(8):1608–1612. doi:10.1016/j.injury.2015.05.035 [CrossRef] PMID:26071324
- Barei DP, Nork SE, Mills WJ, Henley MB, Benirschke SK. Complications associated with internal fixation of high-energy bicondylar tibial plateau fractures utilizing a two-incision technique. J Orthop Trauma. 2004;18(10):649–657. doi:10.1097/00005131-200411000-00001 [CrossRef] PMID:15507817
- Kugelman D, Qatu A, Haglin J, Leucht P, Konda S, Egol K. Complications and unplanned outcomes following operative treatment of tibial plateau fractures. Injury. 2017;48(10):2221–2229. doi:10.1016/j.injury.2017.07.016 [CrossRef] PMID:28733042
- Henry P, Wasserstein D, Paterson M, Kreder H, Jenkinson R. Risk factors for reoperation and mortality after the operative treatment of tibial plateau fractures in Ontario, 1996–2009. J Orthop Trauma. 2015;29(4):182–188. doi:10.1097/BOT.0000000000000237 [CrossRef] PMID:25233159
- Husain A, Pollak AN, Moehring HD, Olson SA, Chapman MW. Removal of intramedullary nails from the femur: a review of 45 cases. J Orthop Trauma. 1996;10(8):560–562. doi:10.1097/00005131-199611000-00009 [CrossRef] PMID:8915919
- Gage MJ, Egol KA. Painful hardware: what to do?Curr Orthop Pract.2014;25(3):198–202. doi:10.1097/BCO.0000000000000108 [CrossRef]
- Minkowitz RB, Bhadsavle S, Walsh M, Egol KA. Removal of painful orthopaedic implants after fracture union. J Bone Joint Surg Am. 2007;89(9):1906–1912. doi:10.2106/JBJS.F.01536 [CrossRef] PMID:17768185
- Garner MR, Thacher RR, Ni A, Berkes MB, Lorich DG. Elective removal of implants after open reduction and internal fixation of tibial plateau fractures improves clinical outcomes. Arch Orthop Trauma Surg. 2015;135(11):1491–1496. doi:10.1007/s00402-015-2299-2 [CrossRef] PMID:26264713
- Reith G, Schmitz-Greven V, Hensel KO, et al. Metal implant removal: benefits and drawbacks—a patient survey. BMC Surg. 2015;15(1):96. doi:10.1186/s12893-015-0081-6 [CrossRef] PMID:26250649
- Busam ML, Esther RJ, Obremskey WT. Hardware removal: indications and expectations. J Am Acad Orthop Surg. 2006;14(2):113–120. doi:10.5435/00124635-200602000-00006 [CrossRef] PMID:16467186
- Hamaker M, O'Hara NN, Eglseder WA, Sciadini MF, Nascone JW, O'Toole RV. Radiographic predictors of symptomatic screw removal after retrograde femoral nail insertion. Injury. 2017;48(3):758–762. doi:10.1016/j.injury.2017.01.015 [CrossRef] PMID:28153480
- Harrell FE Jr., Regression Modeling Strategies with Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. 2nd ed. Cham, NY: Springer; 2015.
- Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol.2007;165(6):710–718. doi:10.1093/aje/kwk052 [CrossRef] PMID:17182981
- Mascarenhas D, Connelly D, O'Hara NN, et al. Radiographic predictors of symptomatic locking screw removal after treatment of tibial fractures with intramedullary nails. Injury. 2018;49(12):2284–2289. doi:10.1016/j.injury.2018.09.026 [CrossRef] PMID:30245279
- Little RJ, Rubin DB. Statistical Analysis with Missing Data. Hoboken, NJ: John Wiley & Sons; 2014.
- Kleinbaum DG, Klein M. Logistic Regression: A Self-Learning Text. 3rd ed. New York, NY: Springer; 2010. doi:10.1007/978-1-4419-1742-3 [CrossRef]
- Greenberg A, Kadar A, Drexler M, et al. Functional outcomes after removal of hardware in patellar fracture: are we helping our patients?Arch Orthop Trauma Surg. 2018;138(3):325–330. doi:10.1007/s00402-017-2852-2 [CrossRef] PMID:29185046
- Kubiak EN, Fulkerson E, Strauss E, Egol KA. The evolution of locked plates. J Bone Joint Surg Am. 2006;88(suppl 4):189–200. PMID:17142448
- Weaver MJ, Harris MB, Strom AC, et al. Fracture pattern and fixation type related to loss of reduction in bicondylar tibial plateau fractures. Injury. 2012;43(6):864–869. doi:10.1016/j.injury.2011.10.035 [CrossRef] PMID:22169068
- Manidakis N, Dosani A, Dimitriou R, Stengel D, Matthews S, Giannoudis P. Tibial plateau fractures: functional outcome and incidence of osteoarthritis in 125 cases. Int Orthop. 2010;34(4):565–570. doi:10.1007/s00264-009-0790-5 [CrossRef] PMID:19440710
- Choo KJ, Morshed S. Postoperative complications after repair of tibial plateau fractures. J Knee Surg. 2014;27(1):11–19. doi:10.1055/s-0033-1363517 [CrossRef] PMID:24343428
- Collinge CA, Koerner JD, Yoon RS, Beltran MJ, Liporace FA. Is there an optimal proximal locking screw length in retrograde intramedullary femoral nailing? Can we stop measuring for these screws?J Orthop Trauma. 2015;29(10):e421–e424. doi:10.1097/BOT.0000000000000353 [CrossRef] PMID:25946415
- Ali AM, El-Shafie M, Willett KM. Failure of fixation of tibial plateau fractures. J Orthop Trauma. 2002;16(5):323–329. doi:10.1097/00005131-200205000-00006 [CrossRef] PMID:11972075
- Ricci WM, Rudzki JR, Borrelli J Jr, . Treatment of complex proximal tibia fractures with the less invasive skeletal stabilization system. J Orthop Trauma. 2004;18(8):521–527. doi:10.1097/00005131-200409000-00007 [CrossRef] PMID:15475847
- Karanicolas PJ, Bhandari M, Taromi B, et al. Blinding of outcomes in trials of orthopaedic trauma: an opportunity to enhance the validity of clinical trials. J Bone Joint Surg Am. 2008;90(5):1026–1033. doi:10.2106/JBJS.G.00963 [CrossRef] PMID:18451395
- Hanson B, van der Werken C, Stengel D. Surgeons' beliefs and perceptions about removal of orthopaedic implants. BMC Musculoskelet Disord. 2008;9(1):73. doi:10.1186/1471-2474-9-73 [CrossRef] PMID:18501014
- Okike K, O'Toole RV, Pollak AN, et al. Survey finds few orthopedic surgeons know the costs of the devices they implant. Health Aff (Millwood). 2014;33(1):103–109. doi:10.1377/hlthaff.2013.0453 [CrossRef] PMID:24395941
- Okike K, Pollak R, O'Toole RV, Pollak AN. “Red-Yellow-Green”: effect of an initiative to guide surgeon choice of orthopaedic implants. J Bone Joint Surg Am. 2017;99(7):e33. doi:10.2106/JBJS.16.00271 [CrossRef] PMID:28375897
Patient Demographics and Injury and Implant Characteristics
|Parameter||Symptomatic Implants (n=82)||Controls (n=132)||P|
|Age, y, mean (95% CI)||44.2 (41.2–47.3)||47.2 (44.6–49.9)||.16|
|Sex, male, No. (%)||42 (51.2)||84 (63.6)||.07|
|Race, No. (%)||.10|
| White||52 (63.4)||87 (65.9)|
| Black||29 (35.4)||36 (27.3)|
| Other or missing||1 (1.2)||9 (6.8)|
|BMI, kg/m2, mean (95% CI)||28.2 (26.6–29.8)||30.1 (28.3–31.8)||.14|
|Smoker, No. (%)||26 (31.7)||51 (38.6)||.33|
|ISS, mean (95% CI)||13.2 (11.3–15.2)||12.0 (10.9–13.1)||.25|
|Bicondylar fracture, No. (%)||60 (73.2)||75 (56.8)||.02a|
|Contralateral LE fracture, No. (%)||21 (25.6)||29 (22.0)||.54|
|Open fracture, No. (%)||4 (4.9)||12 (9.1)||.26|
|Follow-up, mo, mean (95% CI)||23.8 (20.6–27.1)||9.5 (7.0–12.0)||<.01a|
|Time to removal, mo, mean (SD)||18.3 (13.1)|
|Plates, mean (95% CI)||1.4 (1.3–1.5)||1.4 (1.3–1.5)||.73|
|Plate length, mm, mean (95% CI)||201.1 (181.0–221.2)||178.4 (161.9–195.0)||.09|
|Medial plates, No. (%)||29 (35.4)||39 (29.6)||.37|
|Implant distance to joint, mm, mean (95% CI)||3.9 (3.2–4.6)||4.6 (3.9–5.3)||.22|
|Protruding screws per fracture, mean (95% CI)||3.5 (3.1–4.0)||2.4 (2.1–2.8)||<.01a|
|Maximum screw protrusion, mm, mean (95% CI)||3.9 (3.4–4.5)||3.1 (2.7–3.5)||.01a|
Multivariable Prognostic Model for Implant Removal After Tibial Plateau Fixation (n=214)
|Model Parameter||Beta Coefficient||Standard Error||Adjusted OR||95% CI||P|
|Screws protruding, per screw||0.28||0.08||1.32||1.13–1.55||<.001|
| No||Reference (0.00)||Reference (1.00)|
|Age, per 10 y||−0.20||0.11||0.82||0.66–1.01||.06|
|Body mass index, per 5 kg/m2||−0.17||0.10||0.84||0.70–1.01||.05|
| No||Reference (0.00)||Reference (1.00)|