The vast majority of acute nondisplaced scaphoid fractures are currently treated nonoperatively with a cast. Approximately 85% to 90% of these fractures will unite if diagnosed and treated promptly1-4. Traditional cast treatment is reliable and inexpensive and has a low complication rate. The main disadvantages are the long immobilization time required to achieve union, joint stiffness5,6, decreased grip strength, and the delay before the patient can return to work.
An alternative treatment was introduced in 1984 when Herbert and Fisher presented their results with use of a new headless fixation screw7. They used a double-threaded screw with differential pitch at the loading and trailing ends to provide rigid fixation, and thus cast immobilization was rarely required. All acute fractures in their series had united by twelve months. On the basis of their personal experience, the authors also suggested that the true prevalence of nonunion after nonoperative treatment was close to 50%7.
Subsequent studies demonstrated good results, with high union rates, with use of the Herbert bone screw6,8-10. With the apparent advantages of rigid fixation, a high union rate, a short immobilization time, and an early return to work, this method became an attractive alternative to the traditional cast treatment. However, this more complicated method of treatment requires access to operative facilities, is more expensive, and is technically demanding, necessitating a high level of surgical skill. As a result of the complex three-dimensional anatomy of the scaphoid, there are also technical difficulties associated with this surgical technique11. Furthermore, concerns have been raised that the procedure required to insert the implant through the scaphotrapezial joint could result in osteoarthritis12.
The true long-term benefits of internal fixation have not been adequately determined in randomized controlled trials designed to assess the balance between the advantages and disadvantages of the two fundamentally different treatment options. To elucidate this issue, we compared operative fixation of acute scaphoid fractures with nonoperative treatment in a cast in a randomized controlled study of patients followed for eight to thirteen years (median, ten years) after the injury. This report focuses on the long-term outcomes for the seventy-five individuals who were examined clinically and radiographically.
Patients
All patients with a scaphoid fracture who were referred to the Department of Hand Surgery, Uppsala University Hospital, Uppsala, Sweden, between 1992 and 1997 were considered for participation in a randomized study comparing operative treatment with nonoperative treatment in a cast (Fig. 1). The inclusion criteria were an age of seventeen to sixty-five years at the time of injury, a traumatic injury that had occurred less than twenty-eight days before the patient was seen at the Department of Hand Surgery, a radiographically mature skeleton, an isolated scaphoid fracture that was nondisplaced or minimally displaced (by =1 mm) as seen on plain radiographs, and an expectation, on the part of the examiner, that the patient would comply with study instructions. Patients with preexisting symptoms due to a previous known wrist injury and those with a fracture of the tuberosity, coexisting wrist fractures, or a transscaphoid perilunate dislocation were excluded. Patients who agreed to participate were given written information about the risks and benefits of operative compared with nonoperative treatment. Verbal consent was then obtained.
The study was conducted according to the ethical principles of the Helsinki Declaration and approved by our institutional review board (the regional ethical review board of Uppsala University).
All patients were examined clinically and with plain radiography before inclusion in the study. The classification of the fracture as a criterion for inclusion in the study was determined solely with the plain radiographs, as it is in a standard clinical setting. After inclusion and randomization, patients who had been entered into the study were also examined with computed tomography of the injured wrist.
Allocation was based on a table of randomized odd and even numbers generated by a computer. The random numbers were blocked in groups of eight to ensure study groups of approximately the same size. An administrative assistant carried out the allocation process, and the allocations were concealed until interventions were assigned. After giving informed consent and being assigned to a treatment group, the patients who had been allocated to receive the nonoperative regimen were treated by the doctor who had been initially consulted. Patients who were assigned to the operative regimen were scheduled for surgery as soon as practically possible, usually within a week. A total of 245 patients with an injury to the scaphoid bone were assessed for eligibility (Fig. 1); eighty-five were enrolled, and eighty-three received the allocated treatment. One of the eighty-five enrolled patients did not keep the appointment for the surgery, and there was a problem with the planning of the surgery for another patient, who elected to continue nonoperative treatment. Of the 160 excluded patients, 126 did not meet the inclusion criteria: seventeen were too young or too old, forty-five had sustained the fracture more than four weeks before they were seen at the Department of Hand Surgery, thirty-one had a fracture of the tuberosity or distal articular facet, thirteen had coexisting wrist/forearm fractures, nine had a transscaphoid dislocation, eight had a complex dislocated scaphoid injury, one had other serious disease, and two had had a previous injury of the same hand. Six patients refused to participate in the trial, and twenty-eight were excluded for other reasons (three were nonresidents, three did not speak the Swedish language, one was in prison, seven presented with too vague a time history, three could not be scheduled for surgery, and eleven were not included for unclear reasons).
The eighty-five enrolled individuals included thirty-two male and ten female patients allocated to receive nonoperative treatment and thirty-four male and nine female patients allocated to receive operative treatment. The mean age at the time of injury was thirty-one years in the group allocated to receive nonoperative treatment and thirty-two years in the group allocated to receive operative treatment. The injuries involved nineteen right hands and twenty-three left hands in the group allocated to receive nonoperative treatment and twenty right hands and twenty-three left hands in the group allocated to receive operative treatment.
Fifty-one of the eighty-three patients who were enrolled and received the allocated treatment had a nondisplaced fracture. Thirty individuals had small extra fragments at the fracture site, and eleven had a step-off of =1 mm between the two fragments.
Treatment Modes
Nonoperative Treatment
Forty-two patients were allocated to receive nonoperative treatment, which consisted of immobilization in a below-the-elbow scaphoid cast with the thumb held in palmar abduction, the interphalangeal joint free, and the wrist in neutral or slight extension. The cast was initially worn for a planned period of six weeks. It was then removed; radiographs were made; and, if it was judged to be necessary by the treating surgeon, a new cast was applied and worn for another two to four weeks. The surgeon was free to decide on the number of follow-up visits, the intervals between them, and the need for radiographic studies. The surgeon removed the cast when the fracture was considered to be united, on the basis of the clinical findings, standard radiographs, and occasionally computed tomographic scans. Mobilization was then encouraged. Hand therapy training was initiated when indicated. The cast was worn for a mean of ten weeks (range, three to twenty weeks).
Operative Treatment
Surgery was performed on forty-one of the forty-three patients who had been randomized to receive operative treatment. It was done with use of a brachial plexus blockade in thirty-nine patients, general anesthesia in one, and a combination of both in another. Patients were placed in a supine position with the arm on a radiolucent arm-board. A tourniquet was routinely utilized. Twenty-eight fractures were exposed through a volar approach centered over the tubercle of the scaphoid. The flexor carpi radialis tendon sheath was opened, and an arthrotomy was performed through the floor of the tendon sheath. Eleven fractures were approached with a mini-incision that exposed only the scaphotrapezial joint; one, with a dorsal approach; and one, with a combined volar and dorsal approach. A headless bone screw was inserted as described by Herbert and Fisher7; the guiding jig was used for the majority of fractures, but a freehand technique was also used. The distal end of the scaphoid was mobilized and was elevated with use of a periosteal elevator. Alternatively, a trough was created in the anterior aspect of the trapezium with a rongeur to facilitate insertion. The scaphoid fracture was stabilized with a standard Herbert screw (Zimmer, Warsaw, Indiana) in thirty-six patients, and a cannulated Herbert-Whipple screw (Zimmer) was used in two patients. The bone screw was not positioned correctly in two patients; it was therefore removed, and the fracture was treated nonoperatively. One patient who had been allocated to the operative treatment group did not receive the implant as the fracture line was difficult to identify at the time of surgery. On the basis of the intention-to-treat principle, these patients were analyzed as if they were in the operative treatment group.
After skin closure, a well-padded short arm noncircumferential dorsal plaster splint, with the thumb left free, was applied and worn for two weeks. The immobilization time was extended for a few patients because of additional findings or procedures at the time of the surgery (an incomplete ligament injury in two patients and bone-grafting for a comminuted fracture in three). The splint was then changed to a thumb spica cast. The patients in the operatively treated group wore the cast for a mean of three weeks (range, two to sixteen weeks). Two surgeons (B.V. and F.a.E.) performed all operative procedures.
Long-Term Follow-up
At a median of ten years (mean, 10.2 years; range, eight to thirteen years) after the original injury, all patients included in the trial were asked to return for a reexamination. Of the original eighty-three patients who had received the allocated treatment, two had died, leaving eighty-one patients. In addition, six patients did not return for follow-up: two were abroad, three did not reply to our requests, and one refused to return. Thus, seventy-five (93%) of the individuals who were still alive returned for the clinical and radiographic follow-up, and their characteristics are presented in Table I. There were no significant differences between the two study groups with respect to gender, age, dominant or injured hand, circumstances concerning or mechanism of the injury, fracture classification (Table II), or occupation. An unbiased surgeon who had not been involved in the treatment phase of the study performed the clinical examination, which included inspection, evaluation of scar sensibility when applicable, palpation for tenderness, measurement of joint movement with a goniometer, measurement of grip strength with use of the Jamar dynamometer (Sammons Preston, Bolingbrook, Illinois), and measurement of pinch strength with the Baseline Mechanical Pinch Gauge (Fabrication Enterprises, White Plains, New York).
Patients were also requested to answer a questionnaire containing general questions and questions about perceived hand problems as well as two complete limb-specific outcome instruments, the DASH (Disabilities of the Arm, Shoulder and Hand)13 and the PRWE (Patient-Rated Wrist Evaluation)14. The DASH is a thirty-item, self-report questionnaire designed to measure physical function and symptoms in people with any of several musculoskeletal disorders of the upper limb. The overall score can range from 0 (best) to 100 points (worst). The PRWE contains fifteen items with three subscales: pain, specific activities, and usual activities. The results are scored from 0 to 100 points, with higher scores indicating greater pain and disability.
The radiographic examination included standardized images of both wrist joints and scaphoid images of the injured wrist. Posteroanterior and lateral images were made with the wrist in a neutral position (including neutral forearm rotation). Four different scaphoid images were made with the palmar side of the wrist facing the x-ray table: one was made with the x-ray beam tilted 10° in the distal direction; one, with it tilted 10° in the proximal direction; one, with it tilted 20° in the ulnar direction; and one, with it tilted 25° in the radial direction15. Computed tomography scans of both wrists along the long axis of the scaphoid were performed with a Siemens Somatom scanner (Erlangen, Germany) according to a standardized protocol. The protocol was similar to that described by Bain et al.16 except that the wrist was held in a neutral position. Images were acquired with continuous 0.75-mm slices. A wrist immobilization device was routinely utilized. Two patients were examined at their local hospital. For unknown reasons, one patient did not keep the appointment for the computed tomography examination and only the injured wrist was examined in one patient.
Radiographic assessments were made with use of an IMPAX PACS DS 3000 workstation (release 4.5; Agfa-Gevaert Group, Mortsel, Belgium). Osteoarthritis was graded as 1 (a normal joint), 2 (narrowing of the joint space), 3 (osteophytes), 4 (narrowing and osteophytes), or 5 (narrowing, osteophytes, and subchondral sclerosis)17. The radiolunate angle was measured as described by Larsen et al.18,19. Carpal height is defined as the distance from the base of the third metacarpal to the subchondral sclerotic line of the distal radial articular surface as measured along the axis extended from the third metacarpal20,21. The carpal height ratio is obtained by dividing the carpal height by the length of the third metacarpal20,21. The carpal height index corrects for differences in body habitus and is determined by dividing the carpal height ratio of the injured hand by that of the normal hand21,22.
Statistical Analysis
The intention-to-treat principle was utilized in the assessment. Results are descriptively presented as means and medians with 95% confidence intervals. Nominal data were analyzed with the Fisher exact test and the chi-square test where applicable, and interval data were analyzed with the t test if they were considered to be normally distributed. The Mann-Whitney U test was utilized for nonparametric statistics. All analyses were performed with use of standard statistical software.
At the time of the long-term follow-up, all seventy-five fractures were found to have united. However, at three months, one patient in the nonoperatively treated group had a clear gap at the fracture site suggestive of nonunion and was successfully treated with bone-grafting and Herbert bone screw fixation. This was regarded as a complication of treatment.
There were eight complications in the operatively treated group (Table III). Two of them were identified during the operation; in both cases, the bone screw was not positioned satisfactorily and was therefore removed. In one of those cases, the fracture was stabilized with a Kirschner wire and a cast; in the other, the fracture was treated only with a cast. In addition, there were two soft-tissue complications, in two different patients, during the surgery; these consisted of a partial injury of the flexor carpi radialis tendon and a partial injury of the volar part of the scapholunate ligament. These four complications required prolonged cast treatment. Three other patients in the operatively treated group were identified as having malpositioning of the screw with reactive erosion of, or injury to, the opposite joint surface at the ten-year follow-up visit. Finally, there was one nonoperative complication in the operatively treated group; this was a sympathetically maintained pain syndrome, which resolved after intense physical therapy. There were no infections.
At the time of the long-term follow-up, the Herbert screw had been removed from six patients. Five of these removals were due to persistent pain in the area of the previous scaphoid injury. One of these patients stated that the pain disappeared after removal of the implant. Another patient sustained a refracture of the same scaphoid proximal to the Herbert screw, after which a pseudarthrosis developed; this was successfully treated with removal of the screw, bone-grafting, and Kirschner-wire fixation.
Patient Assessments
The two treatment groups had a low mean score on both the DASH (4 points in the nonoperatively treated group and 3 points in the operatively treated group) and the PRWE (6 points in both groups), indicating that most patients were asymptomatic (Table IV). Fifty-seven percent (twenty) of the nonoperatively treated patients had a score of 0 points on the PRWE and 51% (eighteen) had a score of 0 points on the DASH, compared with 45% (eighteen) and 38% (fifteen), respectively, in the operatively treated group (p = 0.36 for the difference in the PRWE scores and p = 0.35 for the difference in the DASH scores). Only two patients had high DASH scores (32 and 49 points). Neither of them had problems related to the previous wrist fracture; one had a painful shoulder, and one had symptoms consistent with a carpal tunnel syndrome.
More individuals in the operatively treated group (fifteen) than in the nonoperatively treated group (eight) reported that the wrist did not function as well as it had prior to the injury, but the difference was not significant (p = 0.21; Table V). The two groups reported similar experiences of weakness in the wrist. The number of individuals who reported limitation of wrist movement was also larger in the operatively treated group (eleven) than in the nonoperatively treated group (six), but again the difference was not significant (p = 0.41).
Objective Assessment
The ranges of motion of the injured wrist were greater in the nonoperatively treated group than they were in the operatively treated group, and the difference in radial deviation was significant (see Appendix). However, when radial deviation was expressed as a ratio of the injured to the uninjured side, the difference between groups was not significant (see Appendix).
Grip strength and pinch strength were also numerically but not significantly greater in the nonoperatively treated group. Eleven patients in the operatively treated group had disturbed sensibility around the scar, but few reported subjective discomfort.
Radiographic Assessment
Osteoarthritis in the scaphotrapezial joint on the injured side was diagnosed with computed tomography in one patient in the nonoperatively treated group and in eleven in the operatively treated group (p = 0.005). Plain radiography revealed osteoarthritis in the scaphotrapezial joint in no patients in the nonoperatively treated group and in five in the operatively treated group (p = 0.032). Computed tomography also revealed signs of osteoarthritis in the radiocarpal joint in three patients in the nonoperatively treated group and in six in the operatively treated group.
There was no significant difference in the radiolunate angle between the patients who had been treated nonoperatively and those who had been operated on (see Appendix). Neither the carpal height ratio nor the carpal height index differed significantly between the groups. Furthermore, the height-to-length ratio measured with computed tomography was similar in the two groups.
Cystic changes of >3 mm were found on plain radiographs of five scaphoids in the operatively treated group compared with none in the nonoperatively treated group (p = 0.057). When assessed with computed tomography, however, six scaphoids were found to have cysts in the nonoperatively treated group and eleven had cysts in the operatively treated group (p = 0.35). Two patients in the operatively treated group had large multicystic changes around the implant, and another patient had a radiolucent zone around the implant suggestive of implant loosening (Figs. 2 and 3).
Malpositioning of the screw, defined as part of the screw protruding outside of the cortical boundaries of the scaphoid, was found in fourteen individuals examined with computed tomography, although most of the protrusions were minor, with the screw probably lying well within the articular cartilage. Seven of these screws protruded distally; four, proximally; and three, both proximally and distally. There was reaction on the opposing bone or joint surface in only three cases.
Ten years after being treated for an acute scaphoid fracture, the patients in this study generally perceived themselves as having good upper-limb and hand function, as indicated by the results of the limb-specific outcome instruments. Only two patients had a high (poor) score on the DASH instrument, and in both cases this was due to problems related to conditions not associated with the original fracture. The DASH scores were similar to those reported by others after shorter time periods following treatment of scaphoid fractures23,24.
The most important observation in this study was a significant increase in the prevalence of osteoarthritis of the scaphotrapezial joint identified with computed tomography and plain radiography in the operatively treated group. Saedén et al.25 made a similar observation, also based on computed tomography assessment, twelve years after internal fixation of scaphoid fractures. As the attrition rate was very low in the present study, our results can actually be seen as a validation of those in the study by Saedén et al., which was limited by the fact that only 61% of the nonoperatively treated patients could be evaluated. The current results are not in agreement with those of Callanan et al.26 and Kehoe et al.27, who used the less sensitive technique of plain radiography. They found no evidence of osteoarthritis in the scaphotrapezial joint five to seven years after insertion of a Herbert bone screw26 or at a mean of eight years after injury27. The most plausible mechanism of such implant-specific osteoarthritis is damage to the articular cartilage of the distal part of the scaphoid, which is an unavoidable part of the surgical procedure, at least when a volar approach is used. It is noteworthy that osteoarthritis developed in three of the eleven patients who had been operated on with a mini-incision, which is the same rate of osteoarthritis as was seen in those who had been operated on with an open approach. Finally, the twelve individuals who were diagnosed with scaphotrapezial osteoarthritis by means of computed tomography did not differ from the remaining sixty-two patients with respect to perceived upper-limb and hand function as assessed with the DASH and PRWE instruments. This finding is consistent with previous reports of a weak relationship between clinical signs and radiographic findings of scaphotrapezial osteoarthritis28,29. However, most patients in the present study were young, and the subjective outcome after an additional period of time is not yet known.
Prolonged cast immobilization may result in joint stiffness5, and this has been mentioned as a factor in favor of operative fixation of scaphoid fractures. Our study did not confirm any negative long-term effect of immobilization with respect to joint stiffness. On the contrary, the nonoperatively treated patients had slightly, although not significantly, better wrist movement and grip strength compared with those treated operatively. It appears that the previously reported30,31 benefit of better wrist movement in operatively treated individuals disappears with time. It also has been suggested that prolonged immobilization delays a patient's return to work. We previously reported that this is only true for those engaged in manual work who cannot perform their duties while wearing a cast32. Manual workers accounted for fewer than one-third of the individuals in the present study.
In general, surgical treatment of fractures provides an opportunity for accurate fracture fixation and to achieve optimal functional results; however, there is the risk of surgical complications, which are not seen in those treated nonoperatively. In the present study, all procedures were performed according to a preset surgical strategy by two surgeons with considerable experience in wrist surgery. In spite of this, there were eight complications in the operatively treated group, seven of which were directly related to the surgical intervention. However, most complications were minor, and some, but not all, could be rectified during the surgical procedure. Of the fourteen patients in whom computed tomography showed that part of the screw had penetrated the cortical boundaries, three had a reaction on the opposing bone or joint surface. This suggests that accurate screw positioning is difficult, as previously reported33.
In one patient with a malpositioned implant, the bone screw was removed and one of the fracture fragments split during the procedure. Ten years later, computed tomography revealed a previously undiagnosed deformity of that scaphoid bone.
An unexpected observation was large multicystic changes surrounding two of the implants (Fig. 2). The Herbert bone screw, which, like most other scaphoid implants, is made of titanium alloy, was originally described as inert with no adverse effects on bone or soft tissue34. Lately, however, more detailed investigations have revealed low-grade inflammatory and fibrotic reactions around titanium implants35,36. This observation may well represent such a reaction. Alternatively, these changes may be caused by micromotion of the implant or the fracture fragments.
The strengths of the present study are the randomized approach and the very high participation rate (93%) at the time of long-term follow-up.
Internal fixation with a headless bone screw is a popular method for treating acute fractures of the scaphoid. Because of the complex anatomic shape of the scaphoid, internal fixation was previously considered to be technically demanding. A minimally invasive technique with cannulated implants has replaced the previously used open approach and has greatly facilitated the procedure. Given the advantage of a high union rate, a short time in a cast, and an early return to work37-39, some authors have recommended this technique for the majority of acute scaphoid fractures40. In view of all of these factors, the question arises as to whether this technique is in reality a conceptual leap in the treatment of scaphoid fractures.
The key issue is whether this new technique should be recommended as a routine treatment for an unselected group of individuals with scaphoid fractures. The best scientific support for a clinical decision is found in the results of a recent systematic review and meta-analysis, which did not demonstrate evidence that operative treatment of a nondisplaced or minimally displaced scaphoid waist fracture provided better outcomes, compared with those of cast treatment, with regard to such factors as the nonunion rate, return to work, grip strength, range of wrist motion, or patient satisfaction41. Operative treatment does, however, cause more complications and perhaps is associated with a higher risk of scaphotrapezial osteoarthritis. The results of the present study fully support the conclusions drawn by Yin et al.41. This study showed that the primary benefit of operative treatment—i.e., a short immobilization time and an early return to work—was transient. Our observation of an increased risk of osteoarthritis in the operatively treated group points to the importance of careful selection of patients who may benefit from operative treatment.
Note: The authors are grateful to Anders Amilon, MD, and Stefan Lewold, MD, PhD, for examining two of the patients at their local hospital and to Jan Blond, MD, for arranging the radiographic examination of one of the patients.