Diagnostic accuracy of diffusion tensor imaging in amyotrophic lateral sclerosis: a systematic review and individual patient data meta-analysis

Abstract

Rationale and objectives: There have been a large number of case-control studies using diffusion tensor imaging (DTI) in amyotrophic lateral sclerosis (ALS). The objective of this study was to perform an individual patient data (IPD) meta-analysis for the estimation of the diagnostic accuracy measures of DTI in the diagnosis of ALS using corticospinal tract data.

Materials and methods: MEDLINE, EMBASE, CINAHL, and Cochrane databases (1966-April 2011) were searched. Studies were included if they used DTI region of interest or tractography techniques to compare mean cerebral corticospinal tract fractional anisotropy values between ALS subjects and healthy controls. Corresponding authors from the identified articles were contacted to collect individual patient data. IPD meta-analysis and meta-regression were performed using Stata. Meta-regression covariate analysis included age, gender, disease duration, and Revised Amyotrophic Lateral Sclerosis Functional Rating Scale scores.

Results: Of 30 identified studies, 11 corresponding authors provided IPD and 221 ALS patients and 187 healthy control subjects were available for study. Pooled area under the receiver operating characteristic curve (AUC) was 0.75 (95% CI: 0.66-0.83), pooled sensitivity was 0.68 (95% CI: 0.62-0.75), and pooled specificity was 0.73 (95% CI: 0.66-0.80). Meta-regression showed no significant differences in pooled AUC for each of the covariates. There was moderate to high heterogeneity of pooled AUC estimates. Study quality was generally high. Data from 19 of the 30 eligible studies were not ascertained, raising possibility of selection bias.

Conclusion: Using corticospinal tract individual patient data, the diagnostic accuracy of DTI appears to lack sufficient discrimination in isolation. Additional research efforts and a multimodal approach that also includes ALS mimics will be required to make neuroimaging a critical component in the workup of ALS.

Keywords: Amyotrophic lateral sclerosis; diagnostic accuracy; diagnostic imaging; diffusion tensor imaging; magnetic resonance imaging; meta-analysis.

Figure 1

Flowchart illustrating selection of studies. DTI, diffusion tensor imaging; ROI, region of interest.

Figure 2

Area under summary receiver operating characteristic (SROC) curve for diffusion tensor imaging fractional anisotropy values. Each circle represents an individual study result. Dashed circle represents 95% prediction interval of summary sensitivity (SENS) and specificity (SPEC). AUC, area under the curve.

Figure 3

Forest plot of area under the curve (AUC) for amyotrophic lateral sclerosis diagnosis using diffusion tensor imaging fractional anisotropy values.

Figure 4

Forest plots of sensitivity (SENS) (a) and specificity (SPEC) (b) for diagnosis of amyotrophic lateral sclerosis using diffusion tensor imaging fractional anisotropy values.

Figure 5

Posttest probabilities after diffusion tensor imaging (DTI) for hypothetical populations with different disease pretest probabilities.

Figure 6

Study quality scores. Graph illustrates study quality based on Quality Assessment of Diagnostic Accuracy Studies criteria, expressed as a percent of studies meeting each criterion.

Figure 7

Assessing publication bias. The funnel plot horizontal axis expresses treatment effect; in this instance, measured by area under the receiver operating characteristic curve (AUC). The vertical axis expresses study size, as measured by standard error (SE). Studies with larger standard errors have a wider confidence interval. The graphed vertical line represents the observed mean AUC and the dashed lines represent the 95% confidence limits for the expected distribution for published studies. The points represent the observed distribution of the published studies. Visual inspection demonstrates the presence of publication bias with several of the studies outside the 95% confidence limits. Further, the plot demonstrates that studies with larger standard errors have lower test performance (ie, AUC).