Cover Page. The handle holds various files of this Leiden University dissertation.

Transcription

1 Cover Page The handle holds various files of this Leiden University dissertation. Author: Rhemrev, Stephanus Jacobus Title: The non-displaced scaphoid fracture : evaluation of diagnostic modalities & conservative treatment Issue Date:

2 Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study Based on: Clinical prediction rule for suspected scaphoid fractures: Injury 2010, 41(10): S.J. Rhemrev F.J.P. Beeres R.H. van Leerdam M. Hogervorst D. Ring

3

4 Introduction The clinical utility of diagnostic tests for triaging suspected scaphoid fractures is hindered by the low prevalence of true fractures, approximately 7% on average [16]. Low prevalence illnesses magnify the shortcomings of diagnostic tests since false positives have more influence as true positives become less common. Research intended to improve our ability to triage suspected scaphoid fractures has focussed primarily on radiological techniques. The perfect radiological diagnostic method is however, elusive and even the best radiological techniques can provide misleading information due to the low prevalence of true fractures amongst suspected scaphoid fractures. Therefore, it becomes necessary to consider methods for increasing the pre-test odds of a true fracture prior to ordering a diagnostic test. To date, the ability of demographic factors, injury characteristics and physical examination findings to help triage suspected scaphoid fractures and thereby increase the pre-test odds of a true fracture prior to diagnostic testing has received limited attention. The objective was to develop a clinical prediction rule that yielded a subset of patients that were more likely to have a scaphoid fracture than others who lacked the subset criteria. Methods Seventy-eight consecutive patients diagnosed with a suspected scaphoid fracture in the Emergency Department were included in a prospective cohort study between April 2004 and January The study was performed in a level-one trauma centre with two Emergency Departments. These 78 patients are a subset of the 100 patients evaluated in a previously published study of magnetic resonance imaging (MRI) and bone scintigraphy for the triage of suspected scaphoid fractures. Twenty-two of those patients did not have the clinical tests described below as the supination and pronation strength meter was available in only one of the emergency departments. Suspected scaphoid fracture was diagnosed on the basis of a tender anatomical snuffbox and pain in the snuffbox when applying axial pressure on the first or second digit, a recent trauma (within 48 hours) and no evidence of a fracture on wrist and scaphoid radiographs. Poly-trauma patients, patients aged <18 years and patients with a bilateral suspected scaphoid fracture were excluded. This prospective study was performed in accordance with the standards of the regional Ethical Committee and all patients gave written informed consent for study inclusion. In the Emergency Department, a number of physical examination manoeuvres were performed. In addition, an MRI scan of the wrist was carried out within 24 hours after the initial presentation and a bone scintigraphy of the wrist was performed between 3 and 5 days after trauma. Patient demographics, history of a prior fracture of either hand or wrist fracture and the injury mechanism were recorded. The history of a prior fracture was recorded to evaluate if this would influence the clinical tests described below. Experienced physicians performed the following tests on both the involved and uninvolved arm 70

5 according to a predefined and standardised method: 1. Inspection of the snuffbox for the presence of ecchymosis or oedema. 2. Flexion and extension of the wrist measured using a hand-held goniometer. 3. Supination and pronation strength measured using a custom device (Leiden, the Netherlands) (Figure 1). 4. Grip strength using a Saehan1 Hydraulic Hand Dynamometer (Masan, Korea). Figure 1. Hydraulic hand dynamometer showing the maximum grip strength (arrow). Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 71

6 Strength measurements were performed with the patient sitting on a height-adjusted chair, with the arm at the side and the elbow flexed, with the forearm, wrist and hand in neutral position. For the supination and pronation strength, the patient had to grasp the handle and twist it in pronation and supination. For the grip strength, the distance between the two bars being squeezed on the dynamometer was adjusted to the size of the patient s hand. For both the injured and uninjured side, the same distance was used. First, the physician explained and demonstrated the appropriate technique after which the patient was allowed to test the technique. Once the patient was familiar with the instruments, we tested the maximum grip in a single best attempt. First, on the uninjured side and then on the injured side. For each measurement, the percentage of the injured side compared with the contra lateral side was calculated. Initial scaphoid radiographs were obtained in three planes: (1) a postero-anterior view with the forearm, wrist and hand in neutral position, (2) an oblique view with the wrist in 10º of supination and maximal ulnar deviation and (3) a true lateral view with the wrist resting in the ulnar position on the X-ray plate. First, all radiographs were judged by the treating resident surgeon in the Emergency Department and a resident radiologist. Subsequently, the consultant trauma surgeon and consultant radiologist judged the radiographs. All four observers had to agree that there was no fracture or malalignment for the patient to be eligible for the study. A 1.5 T MRI scan (Siemens, Erlangen, Germany) was used. The patient lies prone on the scanner couch with the involved hand extended forward palm down over the patient s head. A flexible surface coil was wrapped around the wrist. The MR imaging protocol included coronal T1-weighted turbo spin-echo images with a TR of 450 ms, a TE of 13 ms, a field view of 180 mm, a base resolution of 512, two averages, a slice thickness of 3 mm with a distance factor of 10% (0,3 mm) and a scan time of 2.17 min. The parameters for the coronal fat-suppressed T2-weighted fast spin-echo images were 5220/73 ms (TR/TE), a field of view of 220 mm, a base resolution of 448, three averages, a slice thickness of 3 mm with a distance factor of 10% (0,3 mm) and a scan time of 4.33 min. Bone scintigraphy was performed between 3 and 5 days after trauma, using a standard protocol of images of the early static phase, on a Sky-Light gamma camera (Philips, Eindhoven, the Netherlands). Palmar and dorsal images of both wrists were performed between 2.5 and 4 hours after the injection of 500 MBq of Technetium-99m diphosphonate (Tc-99m-HDP) visualising osteoblast activity. For every patient, both sides were visualised. Patients with a scaphoid fracture on either MR imaging or bone scintigraphy were treated with a scaphoid forearm cast (a below elbow cast incorporating the thumb in the anatomical position as far as the interphalangeal joint, that is, the metacarpophalangeal (MCP) joint flexed at 45-70º and slight abduction and each interphalangeal joint flexed at 10º) for 6 weeks. Patients with other fractures were treated according to local protocols. Patients with no evidence of a fracture on both MR imaging and bone scintigraphy were treated with a supportive bandage and reassessed 2 and 6 weeks after injury. Based on the outcome of MR scanning, bone scintigraphy, clinical and radiographic follow-up, a 72

7 final diagnosis was made after discharge, according to the following reference standard: if MR imaging and bone scintigraphy both showed a fracture, the final diagnosis was: fracture if MR imaging and bone scintigraphy both showed no fracture, the final diagnosis was: no fracture. In case of a discrepancy between MR imaging and bone scintigraphy, both plain radiographs (6 weeks after injury) and physical re-evaluation during follow-up were used to make a final diagnosis: if any clinical signs remained abnormal after 2 weeks (tender anatomic snuffbox or pain in the snuffbox when applying axial pressure on the first or second digit) and/or there was radiographic evidence of a fracture throughout follow-up, the final diagnosis was: fracture if there were no more clinical signs after 2 weeks (no tender anatomic snuffbox and no pain in the snuffbox when applying axial pressure on the first or second digit) or any radiographic evidence of a fracture throughout follow-up, the final diagnosis was: no fracture. Several variables were evaluated to assess their association with scaphoid fractures, including age, sex, dominant hand, previous fracture of either the involved or uninvolved hand or wrist, tender distal radius, oedema, ecchymosis, mechanism of injury, percentage range of motion (ROM), extension, flexion compared with the contra lateral uninjured side, and percentage of strength (pronation, supination and grip) compared with the contra lateral uninjured side. Univariate analysis consisted of two-sample Student s t-test for age and the nonparametric Mann Whitney U-test for ROM variables that were not normally distributed. Fisher s exact test was used to compare the proportion of categorical variables for patients with and without scaphoid fracture. To determine which variables were predictive for a scaphoid fracture, receiver operating characteristic (ROC) curve analysis was used to determine the relationship between sensitivity (true-positive rate) and 1-specificity (false positive rate) with area under the curve (AUC) to measure accuracy of prediction. The Youden index was applied to identify the optimal cut-off threshold values for significant ROM variables [10]. In addition, multiple logistic regression analysis (backward selection method) was used to control for possible confounding and to derive the odds ratios and predicted probabilities of a scaphoid fracture based on combinations of independent predictors [14]. A prediction algorithm was developed based on combinations of multivariate predictors that significantly differentiate between scaphoid fracture and no scaphoid fracture. Statistical analysis was performed using the SPSS software package (version 16.0, SPSS Inc., Chicago, IL, USA). Two tailed values of p < 0.05 were considered statistically significant. Sensitivity, specificity, positive (PPV) and negative predictive values (NPV) were calculated for a scaphoid fracture and for no fracture. Three factors were considered: extension <50% supination strength 10% presence of a previous fracture of either the involved or uninvolved hand or wrist. Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 73

8 A positive test for a fracture was reached if all three factors were present. A positive test for no fracture was reached if all three factors were absent. Results The study cohort of 78 patients included 13 definite scaphoid fractures (17%), according to the reference standard. The mean age for all patients was 41 years (range: 17-84) with 40 males and 38 females. Table 1. Univariate analysis of factors associated with definite scaphoid fracture. Variable Scaphoid fx No scaphoid fx p Value (N=13) (N=65) Age (years), mean ± SD 42 ± ± Range Sex, no. (%) 0.77 Male 06 (15) 34 (85) Female 07 (18) 31 (82) Dominant hand, no. (%) 0.24 Suspected 08 (22) 28 (78) Contralateral 05 (12) 37 (88) Mechanism, no. (%) 0.94 Fall from standing 06 (17) 29 (83) Fall from height 04 (17) 19 (83) Sports 02 (20) 08 (80) Other 01 (10) 09 (90) Previous fracture, no. (%) 0,01* Yes 03 (75) 01 (25) No 10 (14) 64 (86) Swollen, no. (%) 0.53 Yes 10 (20) 41 (80) No 03 (11) 24 (89) Tender distal radius, no. (%) 0.24 Suspected 08 (22) 28 (78) Contralatral 05 (12) 37 (88) Haematoma, no. (%) 0.10 Yes 06 (29) 15 (71) No 07 (12) 50 (88) ROM (% contralateral side) Extension, median (IQR) 08 (0-50) 50 (31-75) <0.01* Flexion, median (IQR) 50 (30-64) 56 (38-79) 0.25 Strength (% contralateral side) Pronation, median (IQR) 04 (2-13) 15 (7-36) <0.01* Supination, median (IQR) 04 (3-10) 20 (11-41) <0.001* Grip, median (IQR) 3 (0-9) 16 (7-35) <0.01* * Statistically significant. ROM = range of motion, IQR = interquartile range. 74

9 Sensitivity (TPR) Supination 10% 45-degree line of nondiscrimination Area under the Curve = % CI = , p < Supination Sensitivity = 85% Specificity = 77% Specificity (FPR) Figure 2. ROC curve for % supination strength compared to the contralateral side. The optimal cut-off point (furthest from the line of nondiscrimination) is supination strength 10%, which has a sensitivity of 0.85 or 85% and a specificity of 77% (a false-positive rate of 0.23 or 23%). The diagnostic accuracy of % supination strength as a predictor of scaphoid fracture is excellent as demonstrated by an AUC value of Table 2. Diagnostic accuracy of range of motion and strength variables. a Continuous variable Area under curve 95% CI p Value Extension <0.01* Flexion Pronation strength <0.01* Supination strength <0.001* Grip strength <0.01* Extension and supination <0.001* strength CI, confidence interval. a AUC values are determined by receiver-operating characteristic (ROC) curve analysis. * Statistically significant. Table 1 summarises demographics, symptoms and ROM variables (% contra lateral side) between patients who had a scaphoid fracture (n = 13) and those with no scaphoid fracture (n = 65). ROC analysis (Figure 2) demonstrated impressive AUC values for all ROM and strength variables, except flexion (Table 2). The scatterplot (Figure 3) of all 78 patients depicts that cut-off values for both supination strength (<10%) and extension (<50%) correctly classifies 11 of 13 scaphoid fracture patients (sensitivity = 85%). Sensitivity and specificity for extension, pronation strength, supination strength and grip strength indicate good diagnostic performance for sensitivity, although clearly Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 75

10 100 Extension (% Contralateral Side) Scaphoid Fx (n = 13) No Scaphoid Fx (n = 65) Cutoff value - Extension Cutoff value - Supination Supination Strength (% Contralateral Side) Figure 3. Scatterplot showing % supination strength and extension relative to the contralateral side for each of the 78 patients. Black circles denote patients with a definite scaphoid fracture and white triangles denote patients without a scaphoid fracture. Data points touching and to the left of the solid vertical line would be classified as having a possible scaphoid fracture using the supination strength cutoff criterion; data points below the dashed horizontal line would be classified as having a scaphoid fracture according to the extension criterion of <50% Sensitivity Specificity 85% 85% 77% 92% 70 69% 65% Percentage % 34% Pronation Strength Extension Supination Strength Grip Strength Figure 4. Bar charts showing the sensitivity (correct classification of the 13 patients with a scaphoid fracture) and specificity (correct classification of the 65 patients with no scaphoid fracture) based on optimal diagnostic criteria for pronation strength ( 10%), extension (<50%), supination strength ( 10%), and grip strength ( 25%). Sensitivity is 85% for both extension and supination strength predictors. Grip strength has a sensitivity of 92% although its specificity is only 34% due to a high false-positive rate. Further analysis indicated that both supination strength and extension provided significant independent predictive information for differentiating between patients with and without a scaphoid fracture. 76

11 Table 3. Prediction algorithm: probability of definite scaphoid fracture based on three multivariate predictors. a. Previous Supination strength Extension Predicted 95% confidence fracture 10% contralateral side <50% contralateral side probability interval Y Y Y Y Y N Y N Y N Y Y Y N N N Y N N N Y N N N a Probabilities are derived from logistic regression using a generalised estimating equation (GEE) approach and are independent of age, sex, grip strength, and mechanism of injury. grip strength produces a high percentage of false positives, as reflected by its low specificity of 34% (Figure 4). Multivariate logistic regression analysis simultaneously tested the variables in Table 1 and revealed three significant predictors of scaphoid fracture: extension <50% (likelihood ratio test = 5.32, odds ratio = 8.6, p = 0.021), supination strength 10% (likelihood ratio test = 9.92, odds ratio = 11.5, p = 0.002) and presence of a previous hand fracture (likelihood ratio test = 5.58, odds ratio = 27.0, p = 0.018). The other covariates tested by multivariate analysis, including age, sex, grip strength and mechanism of injury, were not predictive of scaphoid fracture (all p > 0.20). The primary objective of the statistical analysis was to identify the independent predictors of a scaphoid fracture and to develop a prediction rule or algorithm based on these variables. We used ROC analysis to identify the optimal cut-off values and then confirmed by multivariate logistic regression analysis, the optimal combination of predictors. Since each of the three multivariate predictors is binary, there are eight possible combinations (2 x 2 x 2 = 8) in which an individual patient could be classified. Each combination has an associated probability of a scaphoid fracture as determined by the regression coefficients in the final regression model, along with a corresponding 95% confidence interval (Table 3). For example, if a patient has a previous hand fracture, with supination strength 10% and extension <50% of the contra lateral side, then the probability of a scaphoid fracture is 97% with a precision of this estimate between 77 and 100%. The same patient with no previous hand fracture has a 50% predicted probability of a scaphoid fracture (95% confidence interval: 27-70%). For a patient with none of the three predictors, the estimated probability of a scaphoid fracture is between 0 and 5%. For the three factors (extension <50%, supination strength <10% and presence of a previous fracture of either the involved or uninvolved hand or wrist) the sensitivity, specificity, PPV and NPV were, using a prevalence of 17% for (1) A scaphoid fracture: sensitivity 0.15, specificity 0.98, PPV 0.61 and NPV (2) No scaphoid fracture: sensitivity 0.46, specificity 0.92, PPV 0.54 and NPV Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 77

12 Discussion This analysis demonstrates the feasibility of developing clinical prediction rules to assist in both the management of patients with suspected scaphoid fractures as well as the interpretation of diagnostic tests. The presence or absence of all three factors were highly specific (98 and 92%) and had very good negative predictive values (85 and 89%) indicating that the three factors identified (extension <50%, supination strength 10% and presence of a previous fracture of either the involved or uninvolved hand or wrist), can help with triage and narrow the cohort of patients considered to have a suspected scaphoid fracture. At a minimum, when fewer factors are present, sophisticated diagnostic tests would be ordered and interpreted with care, and operative treatment would be unlikely. With an increasing number of the three factors present, the pre-test odds are higher and the value of additional diagnostic imaging increases. Patients with a history of acute trauma, clinical signs of a scaphoid fracture but no evidence of fracture on plain radiographs have a fracture in approximately 7% of cases. These occult fractures can lead to complications and a delay in treatment increases this risk. A fast and reliable method for triaging suspected scaphoid fractures would help initiate appropriate treatment of true fractures as quickly as possible and allow patients without fractures to resume activities sooner. Bone scintigraphy has been widely used in the diagnosis of scaphoid fractures [3,4,8,11,16,18]. It has a sensitivity of approximately 95% and a specificity of between 60 and 95%. However, it requires intravenous radioactive isotopes and a delay of at least 72 hours after injury. It involves a radiation dose of 4 msv, which is equivalent to 2 years of natural background radiation [5]. CT has been reported to be a useful technique [20]; false-positive results occur when it is used acutely for scaphoid fractures and it is less sensitive than bone scintigraphy [1,19]. It is suggested that MRI has a sensitivity and specificity of approximately 100% [6,7,11,15]. However, a recent study could not prove that acute MR imaging is superior to bone scintigraphy for the detection of suspected scaphoid fractures [5]. The American College of Radiology considers MRI and plain radiographs to be the most appropriate diagnostic tool for imaging acute scaphoid fractures [2]. The Royal College of Radiologists in the United Kingdom gives equal weight to MRI, CT and bone scintigraphy when plain radiographs are negative [17]. There seems to be a lack of consensus regarding the ideal imaging technique for acute scaphoid trauma [12]. Even the best diagnostic test is problematic because of the low prevalence of true fractures amongst suspected scaphoid fractures. Therefore, increasing the pre-test chance of a scaphoid fracture would be helpful. Unfortunately, there is no available prediction rule accessible in literature. Recent attempts have been made to develop a physical test for suspected scaphoid fractures. These studies are useful, but still limited in value [21]. In line with this, the objective of the present study was to attempt to develop a clinical prediction rule that would help determine the patients who would benefit most from more sophisticated imaging techniques. Several things should be kept in mind when interpreting this data. First, our analysis may only be applicable to the patients in our region. For instance, the rate of true scaphoid fractures amongst 78

13 suspected fractures (17%) is higher than most prior series [9,16]. Surprisingly, the presence of a previous fracture was a significant predictor in this study. Due to the low number of patients with a previous fracture, this factor is probably more random than related to the study. One could argue that the prior fracture might imply that these patients are more active, less cautious or even more prone to fracture, but these are merely assumptions with the small number of previous fractures in the study population. Future studies are needed to further evaluate this factor, together with other parameters such as body mass index that have not been considered in this manuscript. Third, the percentage of women with a suspected scaphoid fracture was also relatively high, and gender was not a predictor of true scaphoid fracture. These differences suggest that patients diagnosed with a suspected scaphoid fracture may differ from region to region. Fourth, the lack of availability of the supination pronation strength meter to surgeons in general makes the applicability of this study questionable but of scientific interest. The device was produced specifically for this study and is not commonly available. It has not yet been used and tested on other patients and study conditions nor has it been validated. In addition, the best reference standard for a true scaphoid fracture remains unclear. Finally, we did not evaluate tenderness of the distal pole of the scaphoid or pronation and supination ROM. The goals of determining history and physical examination parameters, which can better define which patients with normal radiographs and scaphoid tenderness require sophisticated tests to rule out a scaphoid fracture, are worth pursuing. However, given the shortcomings of this study, we do not recommend the use of the three specific identified predictors without further study and emphasise that this study does not give definitive answers. Nonetheless, it is clear that this type of analysis can have a substantial influence on the management of the suspected scaphoid fracture and should be pursued in future research. Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 79

14 References 1. Adey L, Souer JS, Lozano-Calderón S, et al. Computed tomography of suspected scaphoid fractures. J Hand Surg Am 2007, 32(1): American College of Radiology. Expert panel on musculoskeletal imaging, appropriateness criteria. Acute hand and wrist trauma. Reston, American College of Radiology Beeres FJP, Hogervorst M, den Hollander P, Rhemrev S. Outcome of routine bone scintigraphy in suspected scaphoid fractures. Injury 2005, 36(10): Beeres FJP, Hogervorst M, Rhemrev SJ, et al. Reliability of bone scintigraphy for suspected scaphoid fractures. Clin Nucl Med 2007, 32(11): Beeres FJP, Rhemrev SJ, den Hollander P, et al. Early magnetic resonance imaging compared with bone scintigraphy in suspected scaphoid fractures. J Bone Joint Surg Br 2008, 90(9): Breitenseher MJ, Metz VM, Gilula LA, et al. Radiographically occult scaphoid fractures: value of MR imaging in detection. Radiology 1997, 203(1): Brydie A, Raby N. Early MRI in the management of clinical scaphoid fracture. Br J Radiol 2003, 76(905): Chakravarty D, Sloan J, Brenchley J. Risk reduction through skeletal scintigraphy as a screening tool in suspected scaphoid fracture: A literature review. Emerg Med J 2002, 19(6): Cooney WP, Dobyns JH, Linscheid RL. Fractures of the scaphoid: a rational approach to management. Clin Orthop Relat Res 1980, 149: Fluss R, Faraggi D, Reiser B. Estimation of the Youden Index and its associated cutoff point. Biomed J 2005, 47(4): Fowler C, Sullivan B, Williams LA, et al. A comparison of bone scintigraphy and MRI in the early diagnosis of the occult scaphoid waist fracture. Skeletal Radiol 1998, 27(12): Groves AM, Cheow H, Balan K, et al. 16-MDCT in the detection of occult wrist fractures: a comparison with skeletal scintigraphy. Am J Roentgenol 2005, 184(5): Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982, 143(1): Hosmer D, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley- Interscience Memarsadeghi M, Breitenseher MJ, Schaefer-Prokop C, et al. Occult scaphoid fractures: comparison of multidetector CT and MR imaging - initial experience. Radiology 2006, 240(1): Ring D, Lozano-Calderón S. Imaging for suspected scaphoid fracture. J Hand Surg Am 2008, 33(6): Royal College of Radiologists. Making the best use of a department of clinical radiology: guidelines for doctors. 5th ed. London, UK: Royal College of Radiologists Thorpe AP, Murray AD, Smith FW, Ferguson J. Clinically suspected scaphoid fracture: A comparison of magnetic resonance imaging and bone scintigraphy. Br J Radiol 1996, 69(818):

15 19. Tiel-van Buul MM, van Beek EJ, Dijkstra PF, et al. Significance of a hot spot on the bone scan after carpal injury - evaluation by computed tomography. Eur J Nucl Med 1993, 20(2): Ty JM, Lozano-Calderón S, Ring D. Computed tomography for triage of suspected scaphoid fractures. Hand (NY) 2008, 3(2): Unay K, Gokcen B, Ozkan K, et al. Examination tests predictive of bone injury in patients with clinically suspected occult scaphoid fracture. Injury 2009, 40(12): Chapter 6 Clinical prediction rule for suspected scaphoid fractures. A prospective cohort study 81