6 Nonvascularized Bone Graft for Scaphoid Nonunion: Predictive Factors
Nonvascularized grafting for scaphoid nonunion has been undertaken for many years, and wide-ranging results have been reported for varied techniques.1 – 4 These techniques include the use of either cancellous or corticocancellous bone taken from various sites, generally the distal radius or the iliac crest.5 This grafting has been supplemented by various methods of fixation ranging from simple impaction of the graft, to K-wires, to screw fixation.6 – 8 Other factors also have an effect on bone healing, including vascularity of the scaphoid fracture fragments, site of the fracture, patient age, smoking, or previous surgery, among others.9 – 13 This chapter collates some of this evidence and allows the surgeon to make a prediction as to the likelihood of surgical success.
In an attempt to understand the place of all variables in the likely success of simple nonvascularized bone grafting and internal fixation we undertook a retrospective review of case records.
A retrospective review of case records identified 159 patients in whom an established nonunion of the scaphoid had been treated by bone grafting and internal fixation. There were 107 patients treated at Wrightington Hospital and 52 patients at Arrowe Park Hospital between April 1991 and February 2003 and between April 1996 and November 2002, respectively. Nonunion was defined as persistence of a fracture gap at least 3 months after the initial trauma, with associated resorption of bone and cystic changes at the fracture site, as seen on the radiographs. There were five different patterns of presentation ( Table 6.1 ).
Patients were included in the study if they had established nonunion of the scaphoid and had undergone nonvascularized bone grafting and internal fixation, with either a screw or Kirschner (K) wires. A total of 19 patients could not remember the month and year of the initial injury, a further nine had less than 6 months follow-up, and six had undergone previous failed surgery for established nonunion. They were thus excluded from the study, as was one patient with a fracture in the coronal plane, leaving 125 patients (126 nonunions) in the study ( Table 6.2 ). The operations were all performed by the authors. A dorsal approach was used in 19 patients with fractures involving the proximal pole. In the remainder, the fracture was approached through an anterior incision. Fibrous tissue between the fragments was excised, and the fracture surfaces were curetted. If a dorsal intercalated segment instability pattern was present, a K-wire was used to correct the deformity. The bone graft was obtained from the iliac crest in 70 patients and from the distal radius in 54. Corticocancellous bone grafts were used when the bone defect was large, and multiple cancellous bone chips when the defect was small. Internal fixation was accomplished using two 1.6 mm K-wires in 46 nonunions and a single screw in 80 cases. Herbert (Zimmer Inc., Warsaw, IN), 3.5 mm (Synthes, Waldenburg, Switzerland), AO and Acutrak screws (Acumed, Beaverton, OR) were used for fixation in 57, 15, and eight cases, respectively. The wrist was immobilized for 8 to 12 weeks after operation. All the patients were followed up clinically and radiologically by one of the two operating surgeons at 6 and 12 weeks, and 6 months, or until union was achieved.
Mean age in years (range)
28 (9 to 59)
Mean time interval from injury to surgery (yrs; range)
4.5 (0.25 to 16)
DISI (radiolunate angle > 10 degrees
Displacement at fracture site
Displaced (> 2 mm)
Abbreviation: DISI, dorsal intercalated segment instability.
The fracture site was described by the fragment ratio, measured using plain radiographs. These radiographs were converted to digital images using a Nikon Coolpix 5000 digital camera (Nikon Corp., Tokyo, Japan) with standard settings. In the digital format, it was easier to identify the margins of the fragments and to make accurate measurements of their size. Because no single view enables visualization of the fracture site in all cases, the long axis of the scaphoid and the fracture are best visualized in the ulnar-deviated and posteroanterior (PA) semipronated oblique views. To calculate the amount of projectional error causing variation in measurement between these two views, the whole length of the scaphoid was measured in 138 films of normal scaphoids, taken consecutively, with standard magnification settings. The estimated standard deviation (SD) within subjects was 0.58 mm, and the intraclass correlation coefficient was 0.932. The repeatability index of 1.6 mm is equivalent to 6.7% of the mean length of the scaphoid. Because the variation in the measured length was not significant, these two views were used to calculate the fragment ratio. The ulnar-deviated PA view was used in 67 nonunions, and the semipronated oblique view in 59.
The bone fragments were measured ( Fig. 6.1 ) using the measurement tool in Adobe Photoshop version 6.0 (Adobe Systems Inc., San Jose, CA). Horizontal lines were drawn at the ends of each fragment to define the length of the fragment. The middle of the fragment was identified and a line was drawn between these points to connect the horizontal lines. The length of this line was measured to determine the size of the fragment. The fragment ratio was then calculated by dividing the proximal fragment size by the sum of the sizes of the proximal and distal fragments. This value was used to describe the site of nonunion. In the 11 vertically oblique fractures, the same procedure was performed but because the line is drawn along the middle of a fragment, it measured the mean length of both.
The outcome variable was defined as bony union or persistent nonunion. Although computed tomographic (CT) and magnetic resonance imaging (MRI) scans are more reliable for assessing bony union, serial radiographs taken at follow-up were used to assess union, as described by Dias in 2001,14 because of the retrospective nature of this study. Bony union was defined as disappearance of the nonunion gap, absence of loosening of the internal fixation, and no displacement of the fragment or graft. Impending nonunion was determined by a persistent gap, loosening of the fixation, or displacement of the fragment. Persistent nonunion was defined when the radiological appearances suggested that the fracture had not united and would not do so without further intervention. The outcome variable was defined as bony union or persistent nonunion.
Intraobserver variability in defining the fracture site was assessed. Three independent blinded observers were asked to measure the fragment ratios for all 126 nonunions. An intra-class correlation coefficient and a repeatability coefficient were calculated in Statsdirect 2.4.4 (Stats Direct Ltd, Cheshire, England) using the Bland Altman method. Univariate logistic regression analyses were performed using each of 15 possible explanatory variables: age, gender, hand dominance, the site of nonunion, the fracture pattern, displacement at the nonunion site, the presence of dorsal intercalated segment instability, the presence of osteoarthritis, previous surgical fixation of an acute fracture, the time interval to surgery, the surgeon, smoking, initial conservative treatment, the method of fixation, and the type of graft. Data on smoking were only recorded in 94 patients. We did not examine vascularity because we believe that its assessment is both subjective and controversial. All logistic regression analyses were performed using SPSS 10.1 (Statistical Package for the Social Sciences SPSS Inc., Chicago, IL) with significance being achieved of p < 0.05 in all instances.
Multivariate logistic regression analyses were also performed. All 15 explanatory variables were initially included, and a stepwise procedure was used to find a subset of variables that could be combined in a model to predict union. The model’s performance in predicting the probability of union was measured using receiver operating characteristic (ROC) analysis in Statsdirect. The area under the curve was used as a measure of predicting the probability of union, where it ranges from 0.5 (no better than chance) to 1.0 (perfection), as well as the sensitivity, the true positive rate of union and specificity, and the true negative rate of union.
Number of Fractures
Number United (%)