This review compared imaging modalities for diagnosis of scaphoid fracture and found bone scintigraphy and magnetic resonance imaging (MRI) useful for excluding fracture and MRI better for confirming fracture. More studies of computed tomography were needed. These conclusions were representative of the data presented, but should be interpreted cautiously in view of potential missed studies and some limitations in analyses.

To compare the diagnostic performance of bone scintigraphy, magnetic resonance imaging (MRI), and computed tomography (CT) for suspected scaphoid fractures.

PubMed and EMBASE were searched from inception to October 2008. Search terms (including methodological terms for test accuracy studies) were reported. Only studies published as full reports in English were included. Bibliographies of included studies were screened for additional articles.

Studies that assessed the diagnostic performance of bone scintigraphy, MRI, CT or ultrasound for suspected scaphoid fractures were eligible for inclusion. Included studies had to use follow-up imaging (radiograph, CT, MRI or bone scintigraphy) or clinical follow-up and/or combined images as the reference standard to confirm diagnosis. Included studies had to report sufficient data to populate 2x2 contingency tables (numbers of true positive, false negative, false positive and true negative test results).

Included studies were conducted between 1981 and 2008. Where reported, mean age of study participants ranged from 22 to 44 years and the period between injury and test ranged from zero to 60 days. Duration of clinical follow-up and timing of follow-up imaging varied between studies and ranged between 10 days and 12 months. Technical details varied within imaging test type and were reported in a table of study details.

Two reviewers independently assessed studies for inclusion. Disagreements were resolved by consensus.

Methodological quality of included studies was assessed using QUADAS criteria. The item "was the period between index test and reference standard short enough" was omitted because follow-up was used as the reference standard. Studies were considered high quality when they scored positive for at least four criteria from: prospective design with consecutive recruitment; appropriate reference standard; avoidance of partial verification; avoidance of differential verification; and interpretation of reference standard without knowledge of index test results. Other studies were considered to be low quality.

Two reviewers independently assessed study quality. Disagreements were resolved by consensus.

Data were extracted to populate 2x2 contingency tables for each index test. Estimates of sensitivity, specificity and the natural logarithm of the diagnostic odds ratio (lnDOR), with 95% confidence intervals (CIs) were calculated for each data set.

Two reviewers independently extracted data. Disagreements were resolved by consensus.

Summary estimates of sensitivity, specificity and lnDOR, with 95% CIs, were calculated using a random-effects model. A correction factor of 0.5 was applied to all cells in the 2x2 table for studies with one or more zero values. Summary likelihood ratios (LRs) were calculated from pooled estimates of sensitivity and specificity, and used to estimate post-test probability.

Summary receiver operating characteristic (SROC) models were constructed for each test using the Moses, Shapiro, Littenberg model (unweighted). Studies of ultrasound were not pooled and were presented separately in a narrative synthesis.

Statistical heterogeneity was assessed using Cochran's Q test and quantified using the I² test.

Between-study heterogeneity was explored using univariable meta-regression analyses. Where there were 10 or more included studies, the following covariates were assessed: sample size (≤50 versus >50); prevalence of scaphoid fracture (≤15% versus >15%); time between injury and index test (≤10 days versus >10 days); study quality (high versus low). Subgroup analyses were conducted for any significant covariate (p< 0.1). Sensitivity analyses were performed by excluding outlier studies.

To compare overall accuracy between imaging modalities, a dummy variable for imaging modality was included in the meta-regression model and SROC model; a p value of less than 0.05 for the regression coefficient of this variable was taken to indicate significant difference.

Twenty-six studies (total number of participants unclear) were included in the review; some studies assessed more than one imaging modality. Most studies had some methodological limitations; only two studies satisfied all criteria. Most studies (>70%) met criteria for adequate patient spectrum, avoidance of partial verification, independent reference test, adequate description of index test, blind assessment of index test, availability of clinical data and handling of uninterpretable results and withdrawals. Few studies (≤35%) met criteria for adequate description of reference standard and blind interpretation of reference standard. The prevalence of scaphoid fracture in included studies ranged from 5% to 50%.

Fifteen studies of bone scintigraphy (1,102 participants) gave a pooled estimate for sensitivity of 97% (95% CI 93% to 99%), a pooled estimate for specificity of 89% (95% CI 83% to 94%) and a pooled estimate for lnDOR of 4.78 (95% CI 4.02 to 5.54). There was significant statistical heterogeneity in estimates of sensitivity (I²=85%).

Ten studies of MRI (513 participants) gave a pooled estimate for sensitivity of 96% (95% CI 91% to 99%), a pooled estimate for specificity of 99% (95% CI 96% to 100%) and a pooled estimate for lnDOR of 6.60 (95% CI 5.43 to 7.76).

Six studies of CT (211 participants) gave a pooled estimate for sensitivity of 93% (95% CI 83% to 98%), a pooled estimate for specificity of 99% (95% CI 96% to 100%) and a pooled estimate for lnDOR of 6.11 (95% CI 4.56 to 7.66).

Two studies reported data for ultrasound (n=72), with sensitivity estimates of 100% (95% CI 48% to 100%) and 78% (95% CI 40% to 97%) and specificity estimates of 98% (95% CI 89% to 100%) and 89% (95% CI 52% to 100%).

Comparative SROC analyses showed that MRI had a higher diagnostic performance than bone scintigraphy (relative DOR 4.85, 95% CI 1.42 to 16.52) and a similar diagnostic performance to CT; there was no significant difference in performance between CT and bone scintigraphy. Meta-regression analyses indicated that longer time periods between injury and imaging were associated with increased specificity for bone scintigraphy; no other variables were significant. Sensitivity analyses did not significantly change results.

Likelihood ratios were reported.

Bone scintigraphy and MRI had high sensitivity and were useful for excluding scaphoid fracture. MRI was more specific and better for confirming fracture. Additional studies were needed to assess the performance of CT.

The review addressed a clearly stated question supported by appropriate inclusion criteria. Restriction of the search to published English-language studies raised the possibility of language and publication biases (acknowledged by the authors). Use of methodological search terms for test accuracy studies was likely to further limit the sensitivity of searches. Therefore, relevant studies may have been omitted from the review. The review process employed rigorous methods to minimise potential for error and/or bias. Some details of included studies and individual study results were reported. Sample sizes were generally small (acknowledged by the authors) and confidence intervals around pooled estimates were consequently wide. Meta-analyses used were broadly appropriate (although bivariate/hierarchical SROC models are now recommended in preference to the Moses and Littenberg model). The review included a number of studies that assessed more than one imaging modality, which thus provided direct comparisons in the same population; however, no direct comparison data were presented. Overall, the authors conclusions were representative of data presented, but should be interpreted cautiously, in view of the potential for missed studies and limitations of the analyses.

Practice: The authors stated that MRI was highly accurate for confirmation and exclusion of scaphoid fracture and might be used as the first choice imaging modality. Bone scintigraphy was inappropriate for confirming scaphoid fractures. The authors stated that the prevalence of abnormality can rarely be generalised beyond the study and that their data should be viewed with caution in respect of generalisability.

Research: The authors stated that more studies were needed to assess diagnostic performance of CT (especially paired design or randomised controlled trials that compared CT with MRI or bone scintigraphy).

Not stated.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.