omech and EBM_Figure 1C.gif
Fig. 1 Periprosthetic supracondylar femur fracture with five bicortical locked screws proximal to the fracture and four locking screws in the distal metaphysis. Lateral postoperative radiographs demonstrating a supracondylar femoral nonunion 8 months after initial surgery.
Courtesy of Greg A. Brown, MD, PhD


Published 12/1/2010
Greg A. Brown, MD, PhD

Why we need Level V evidence

Orthopaedic biomechanics research in the era of evidenced-based medicine

In the current evidence-based medicine (EBM) environment, the emphasis is on Level I research, which is sorely lacking in orthopaedics. Both expert opinion and basic science research are classified as Level V evidence, the lowest level. But they couldn’t be farther apart in their appropriateness to orthopaedics.

Expert opinion is limited by the expert’s biases. Biomechanics basic science research is limited both by assumptions (loading conditions, animal models) and the model’s applicability to human patient subgroups.

Orthopaedic biomechanics can be divided into the following two broad categories:

  • failure analysis
  • the design, development, characterization, or comparison of orthopaedic treatments, procedures, and implants

Failure analysis
Failure analysis applies scientific disciplines such as engineering, metallurgy, biology, or epidemiology so that surgeons can “learn” from failures. These lessons provide the insight needed to improve surgical treatments, procedures, and implants.

To validate the biomechanical hypothesis derived from the failure analysis, researchers must develop a failure model (animal, cadaver, or clinical) that reproduces the relevant failure mode. Therefore, failure analyses and failure models are prerequisites for the second category of biomechanics research.

The significance of this research can be seen in the Journal of Orthopaedic Trauma’s “Leading 20 at 20: Top Cited Articles and Authors in the Journal of Orthopaedic Trauma, 1987–2007” (January 2010 issue). One-fourth of the articles listed were categorized as basic science articles. In three articles, animal models were used to assess bone healing and bone blood flow (failure mode: delayed unions and nonunions; failure model: animals). In two articles, ex vivo models were used (failure modes: screw pullout and proximal femoral bursting; failure models: synthetic bone and cadaver femora; respectively).

In orthopaedic trauma, the presumption seems to be that stiffer and stronger are better. This can be understood for primary bone healing that requires anatomic reduction and rigid fixation. Secondary bone healing, however, requires adequate alignment and relative stability.

A 2007 study presented animal data supporting the fact that secondary bone healing callus is stronger with optimal fracture/construct stability (stiffness). This is not new information. Historically, double-stacked external fixators were too stiff and led to delayed unions or nonunions.

The same lessons are being relearned with locking plates (internal fixators). When multiple locking screws proximal and distal to the fracture stiffen the construct, the result can be a nonunion (Fig. 1). Methods to reduce the stiffness of locking plates are being evaluated. Because ex vivo cadaver studies cannot incorporate the biology of fracture healing, cadaver studies characterizing and comparing implants and constructs can never determine the optimal stiffness for a given fracture configuration.

Fatigue failure
Fatigue failure is another mode of implant/construct failure. Fatigue failure occurs with repeated loading below the construct’s ultimate failure strength. Fatigue strength is measured by using cyclic testing and is reported as the number of cycles to failure for a given load.

Trauma implants will fail, given a sufficient number of loading cycles above their endurance limit. Because fracture healing is a race between bone union and implant failure, fatigue failure typically occurs with delayed unions or nonunions (Fig. 2).

Because fatigue failure depends on the number of load cycles and the biologic timing of fracture healing, ex vivo cadaver studies with cyclic loading are of limited value. Fracture-healing biology cannot be incorporated in a cadaver model. Cadaver studies are worst-case scenarios without fracture healing, and time to failure can be estimated by the number of cycles to failure divided by the number of gait or loading cycles per day.

omech and EBM_Figure 1C.gif
Fig. 1 Periprosthetic supracondylar femur fracture with five bicortical locked screws proximal to the fracture and four locking screws in the distal metaphysis. Lateral postoperative radiographs demonstrating a supracondylar femoral nonunion 8 months after initial surgery.
Courtesy of Greg A. Brown, MD, PhD
omech and EBM_Figure 2A.gif
Fig. 2 Anteroposterior radiograph showing fatigue failure of proximal tibia locking plate with proximal tibia nonunion.
Courtesy of Greg A. Brown, MD, PhD
omech and EBM_Figure 3A.gif
Fig. 3 Anteroposterior radiograph showing failure of proximal humerus locking plate in elderly osteoporotic female with humeral head collapse and screw penetration of humeral head.
Courtesy of Greg A. Brown, MD, PhD

Fixation failure
Multiple failure modes exist for fracture fixation. If a nonunion develops, the implant may fail due to plate breakage, screw breakage, or any other implant failure. Fixation failure can also occur with screw pullout, screw cutout, or screw joint penetration.

Selecting the proper failure mode and failure model is critical to applying failure analysis. A mismatch between the biomechanics research and the clinical failure mode generates confusing information. When proximal humerus locking plates were initially released for clinical use, for example, a video showed a locking plate being pulled out of an apple, which served as the model for the humeral head. The apple split before the locking screws could be pulled out.

This video was made before orthopaedic surgeons understood that proximal humerus locking plates fail with humeral head varus collapse and articular screw penetration. In severely osteoporotic bone, the locking screws do not support the humeral head, and head collapse results in articular screw penetration (Fig. 3). The video used an incorrect failure mode and provided orthopaedic surgeons who used proximal humerus locking plates in severely osteoporotic bone with a false sense of security.

Bringing EBM to biomechanics research
High-quality research requires a hypothesis. In biomechanics research, that means identifying a clinically relevant problem and failure mode and testing the hypothesis through failure analysis. Unfortunately, this research approach is frequently not followed.

Consider the development and release of a new implant. The construct stiffness and cyclic failure strength are measured with a mechanical testing machine, the new implant is often compared to one or more existing implants, and the results that the new device is stiffer (and therefore more stable) and stronger than existing implants are sent for publication.

With proximal humerus fractures, for example, a manuscript may begin with the statement that “proximal humerus fractures are common.” Bringing EBM to bear, the reader might ask: What proximal humerus fracture pattern is being studied? How common are 2-part anatomic neck proximal humerus fractures compared to 2-part surgical neck fractures, 3-part fractures, or 4-part fractures? What failure mode is being studied: varus collapse, articular screw penetration, plate fracture, or metaphyseal screw pullout? Does the selected model fail by the failure mode being studied?

A published biomechanics article that compares implants and constructs, but does not reproduce clinically relevant failure modes misleads orthopaedic surgeons and is incapable of providing high-quality evidence for clinical outcomes.

The biomechanics research regarding tip apex distance in intertrochanteric hip fractures is an excellent example for developing and testing a hypothesis. The tip-apex distance is the sum of the distance from the top of the lag screw to the apex of the femoral head on anteroposterior and lateral radiographs. The hypothesis was that an increasing tip apex distance was associated with a higher rate of fixation failure with lag screw cutout. The failure mode was lag screw cutout. The failure model was an epidemiologic model correlating the tip apex distance with clinical fixation failure and lag screw cutout. Lag screw cutout failure did not occur in any patient with a tip apex distance of 25 mm or less.

The conclusion—that the tip apex distance is an appropriate biomechanical measure—was confirmed with a controlled trial using surgeon education as the intervention. This biomechanics failure analysis provided orthopaedic surgeons with valuable intraoperative measures to reduce fixation failures (cutouts) and improve clinical outcomes.

Balancing basic science and clinical outcomes
The clinical value of biomechanics research is frequently limited because the research cannot assess the clinical outcomes of treatments, procedures, and implants. The ultimate clinical outcome for a given patient depends on a multitude of factors, including patient factors (comorbidities, bone density, prefracture function, treatment adherence, soft-tissue injury), surgeon factors (surgical technique, surgical volume, surgical experience), and implant characteristics (ease of use, material properties, screw/construct configuration).

The nearly infinite number of factor combinations cannot be assessed in any given biomechanics model. Therefore, positive biomechanics evidence is a prerequisite to clinical trials, but clinical trials are needed to balance the above factors between treatment groups to assess clinical outcome differences.

The same issue of the Journal of Orthopaedic Trauma that included the “Leading 20 at 20” article also had five articles dealing with patient-reported functional and quality-of-life outcomes for distal tibia/ankle fractures and syndesmotic injuries. This epitomizes the new era of EBM. In this new era, the contribution of orthopaedic biomechanics research will be to explain failure modes through failure analysis; to develop failure models that will allow the development of improved orthopaedic treatments, procedures, and implants; and to generate hypotheses that clinical trials can test with patient-reported quality-of-life and functional outcomes.

Greg A. Brown, MD, PhD, is a member of the AAOS Guidelines and Technology Oversight Committee. He can be reached at