Download presentation
Presentation is loading. Please wait.
Published bySteven Fowler Modified over 8 years ago
1
Physical examination for diagnosing disc herniation in patients with back pain: systematic review of diagnostic accuracy studies Daniëlle van der Windt and many others…. Primary Care Musculoskeletal Research Centre, Keele University EMGO Instituut and Department of General Practice, VU University medical centre, Amsterdam
2
Content Objectives of the review Writing the protocol: decisions Difficulties & lessons learnt o searching & selecting papers o quality assessment o data extraction o data synthesis Preliminary results Discussion
3
Cochrane reviews of diagnostic accuracy April 2003: Cochrane Steering Group decides to include reviews of diagnostic accuracy studies Cochrane Diagnostic Reviewers Handbook: 2006 - 2008 Pilot reviews (n = 11) Specific software has been developed Publication of first protocols (2007) and reviews 2008?
4
Pilot review (Cochrane Back Review Group) Physical examination for the diagnosis of lumbar radiculopathy due to disc herniation in patients with low-back pain Herniated intervertebral disc: cause of LBP < 2% of patients Accurate clinical diagnosis of lumbar radiculopathy is desirable to avoid unnecessary (surgical) intervention Physical examination decision making process o referral for diagnostic imaging? o suitable candidate for surgery?
5
Objectives To assess the diagnostic performance of tests performed during physical examination in the identification of disc herniation in patients with low-back pain o single tests o combinations of tests To assess the influence of sources of heterogeneity on diagnostic performance
6
Writing the protocol - decisions Definition of target condition o disc herniation (bulging, prolapse, extrusion, …) o lumbar radiculopathy o other specific causes Selection of the reference standard o imaging o surgical findings o recovery following surgery Physical examination for … o identifying presence of disc herniation (yes/no) o identifying level of disc herniation
7
Searching & selecting papers Definition of search strategy based on research question Definition of selection criteria Diagnostic studies are not easy to identify Use of a methodology filter is not recommended (Doust et al. 2005; Leeflang et al. 2006) This review: use of a sensitive filter in Medline & Embase
8
Search results Results in April 2007 (+ update April 2008): o Medline: 1529 hits+ 132 o Embase: 793 hits+ 82 o Cinahl: 321 hits+ 88 Additional hits without use of methodology filter: o Medline: 336 4 papers (< 1969) potentially relevant o Embase: 126 1 paper potentially relevant
9
Quality assessment QUADAS tool (Whiting 2004): o 11 core items o 3 additonal items, relevant to this review - treatment between index and reference test - definition of cut-off points (positivity criteria) - interobserver variation Two assessors, 3 rd in case of disagreement
10
Assessing quality: difficulties Disagreements: 29% of all scored items Some items seemed irrelevant to our review: o Spectrum of patients representative of those who will receive the test in practice? rephrased: “Sample is representative of target population” Some items are strongly related: o Reference test assessed independent of index test? o Results of reference test interpreted without knowledge of index test results?
11
Quality assessment: results
12
Data extraction Characteristics of participants: setting, age, gender, history of low back pain Test characteristics (including cut-offs / positivity criteria) Study design Results: diagnostic 2x2 table
13
Extracting data: difficulties Poor reporting of essential aspects of study design: o patient sampling o reference standard o sequence of testing o who performed which test? Diagnostic 2x2 table not always presented, or impossible to reconstruct Poor reporting of results: even incorrect computations of diagnostic accuracy
14
Results 18 studies 2 case-control studies 16 cohort studies, 5 with a retrospective design Reference test: o surgical findings: 10 studies o diagnostic imaging (MRI, CT): 6 studies o case-control studies: surgery for cases, MRI/CT voor controls More than 10 different types of tests
15
surgical findings (reference test) pos neg a b a + b SLR + (true positive) (false positive) c d c + d SLR - (fase negative) (true negative) a + c b + d N Sensitivity: a / a + c probability of positive test result when diseased Specificity: d / b+d probability of negative test results when healthy Data extraction
16
Results: example Scoliosis
17
Results: example MRI CC Impaired tendon reflexes
18
Straight leg raising test CC Reference test: surgery Pooled estimates (bivariate): - sensitivity: 0.92 (0.87 - 0.95) - specificity: 0.28 (0.18 - 0.40)
19
Straight leg raising test Reference test: MRI/CT CC Reference test: surgery
20
Crossed straight leg raising test CC Crossed straight leg raising test Pooled estimates (bivariate): - sensitivity: 0.23 (0.21 - 0.25) - specificity: 0.89 (0.86 - 0.91)
21
Exploring sources of heterogeneity SLR: diagnostic performance may be poorer cohort studies compared to case control studies exclusion of patients with previous back surgery Limited information on important study characteristics Number of studies for most tests very small
22
Discussion Most studies address individual tests only combination of multiple tests (patient history & PE) more relevant, and possibly more accurate diagnostic models / prediction rules No optimal reference standard to determine if pain is caused by disc herniation Very few studies in primary care populations Diagnostic performance of most tests used during physical examination tests is poor
23
Questions SLR: sensitive, but not very specific suitable for ex cluding disc herniation? (few false negatives) SnNout Crossed SLR: specific, but not very sensitive suitable for in cluding disc herniation? (few false positives) SpPin Q1: Can we always rely on the SpPin & SnNout rules?
24
Questions CC Reference test: surgery Q2: What is more informative? - pooled sensitivity & specificity - Summary ROC curves
25
Ingrid Riphagen (search strategy) Manu Simons & Bert Aertgeerts (selection) Carlo Ammendolia & Arianne Verhagen (quality assessment) Mark Laslett (data extraction) Rick Deyo, Vicky Pennick, Lex Bouter, Riekie de Vet Peer reviewers (protocol & review) Acknowledgments
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.