[

research report

]

Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Uni of Sydney Cumberland on November 24, 2016. For personal use only. No other uses without permission. Copyright © 2015 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

ROBIN HASKINS, BPhty (Hons)1 • PETER G. OSMOTHERLY, PhD1 • DARREN A. RIVETT, PhD1

Diagnostic Clinical Prediction Rules for Specific Subtypes of Low Back Pain: A Systematic Review

P

atients with low back pain (LBP) are generally considered to consist of smaller subgroups that differ meaningfully with regard to their symptomatology, prognosis, and response to various treatments.52 It is hypothesized that such heterogeneity within the LBP patient population may contribute to the relatively modest effect sizes generally observed in most high-quality clinical

trials.4,28 The identification of LBP subgroups has been a research priority for several years,9,21,27,40 and several classification approaches have been TTSTUDY DESIGN: Systematic review.

proposed.49,53,68,77 Traditionally, such classification approaches have been predominantly based on expert opinion and biologic plausibility, with notably little methodological standards for the development of CPRs.

TTOBJECTIVES: To identify diagnostic clinical

TTRESULTS: Of 10 014 studies screened for eligibility,

prediction rules (CPRs) for low back pain (LBP) and to assess their readiness for clinical application.

TTBACKGROUND: Significant research has been

invested into the development of CPRs that may assist in the meaningful subgrouping of patients with LBP. To date, very little is known about diagnostic forms of CPRs for LBP, which relate to the present status or classification of an individual, and whether they have been developed sufficiently to enable their application in clinical practice.

TTMETHODS: A sensitive electronic search strat-

egy using 7 databases was combined with hand searching and citation tracking to identify eligible studies. Two independent reviewers identified relevant studies for inclusion using a 2-stage selection process. The quality appraisal of included studies was conducted by 2 independent raters using the Quality Assessment of Diagnostic Accuracy Studies-2 and checklists composed of accepted

the search identified that 13 diagnostic CPRs for LBP have been derived. Among those, 1 tool for identifying lumbar spinal stenosis and 2 tools for identifying inflammatory back pain have undergone validation. No impact analysis studies were identified.

TTCONCLUSION: Most diagnostic CPRs for LBP

are in their initial development phase and cannot be recommended for use in clinical practice at this time. Validation and impact analysis of the diagnostic CPRs identified in this review are warranted, particularly for those tools that meet an identified unmet need of clinicians who manage patients with LBP.

TTLEVEL OF EVIDENCE: Diagnosis, level 2a–. J Orthop Sports Phys Ther 2015;45(2):61-76. Epub 8 Jan 2015. doi:10.2519/jospt.2015.5723

TTKEY WORDS: decision support techniques, sensitivity, specificity

concordance among them. More recently, there has been greater focus on empirically derived subgrouping methods, including the development of clinical prediction rules (CPRs).6,29 A CPR is a clinical tool designed to assist decision making for individual patients by combining elements from the history, physical examination, and other investigations to make predictions regarding a patient’s diagnosis, prognosis, or likely response to a particular treatment.6,18,69 A CPR is initially derived using multivariable statistical procedures to identify which aspects of a patient’s presentation are independently related to a certain diagnosis or outcome. The tool then undergoes a process of validation, whereby it is applied in new groups of patients in different settings to evaluate its ability to accurately predict that same diagnosis/ outcome. Validated CPRs subsequently undergo impact analysis, by which they are tested to determine whether their clinical application leads to improved patient outcomes or efficiencies in resource consumption.69,70,98 The stage of CPR development has important implications regarding the CPR’s appropriateness to be applied in clinical practice. A CPR that has not undergone validation is not recommended for use in practice, as it may re-

School of Health Sciences, The University of Newcastle, Callaghan, Australia. The authors certify that they have no affiliations with or financial involvement in any organization or entity with a direct financial interest in the subject matter or materials discussed in the article. Address correspondence to Robin Haskins, School of Health Sciences, The Hunter Building, The University of Newcastle, Callaghan, NSW 2308 Australia. E-mail: [email protected] t Copyright ©2015 Journal of Orthopaedic & Sports Physical Therapy® 1

journal of orthopaedic & sports physical therapy | volume 45 | number 2 | february 2015 | 61

45-02 Haskins.indd 61

1/22/2015 10:44:22 AM

[ TABLE 1

research report

Study Eligibility Criteria

1. R  eports on the derivation, validation, and/or impact analysis of 1 or more diagnostic clinical prediction rules related to the nonsurgical management of adults with low back pain 2. T he clinical prediction rule under development contains 2 or more predictor variables 3. T he clinical prediction rule under development was initiated by a formal derivation process in which a larger pool of candidate predictor variables was refined to a smaller set of variables based on their identified independent predictive value using formal multivariable statistical procedures 4. A  tool is clearly presented in sufficient detail that may be applied by a clinician to inform a diagnosis for an individual patient 5. P  ublished in English Exclusion criteria 1. L imited to the investigation of modifiable and/or determinant predictor variables 2. C  linical prediction rule not capable of directly contributing to patient care

Screening

Identification

3. C  onference proceedings/abstracts, dissertations, commentaries, reviews, editorials, letters, study protocols, n = 1 designs (case reports), books, book chapters, book reviews, practice guidelines

Records identified through database search, n = 12 347

Additional records identified through other sources, n = 34

Records after duplicates removed, n = 9980

Titles and abstracts screened, n = 10 014 Records excluded, n = 9863 Full-text articles assessed for eligibility, n = 151 Full-text articles excluded, n = 136 • Not a CPR as defined in this review, n = 86 • Not related to diagnosis of adults with LBP, n = 41 • Ineligible article type, n = 8 • Surgery, n = 1

Eligibility Included

Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Uni of Sydney Cumberland on November 24, 2016. For personal use only. No other uses without permission. Copyright © 2015 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

Inclusion criteria

Studies included in qualitative synthesis, n = 15

FIGURE. PRISMA flow chart. Abbreviations: CPR, clinical prediction rule; LBP, low back pain.

flect chance statistical associations or be specific to the patient sample or setting

from which it was derived.70 It is generally accepted that a CPR that has been

] prospectively validated in new patient cohorts across broad clinical settings may be applied in practice in similar patient populations with confidence in its known predictive accuracy.69,70 However, an important consideration is that, though its predictive accuracy (the amount of agreement between the results from an index test and those from a reference standard)10 may be known, it should not be assumed that a validated CPR’s accuracy will be superior to unassisted clinician judgment, or that application of the rule will result in beneficial clinical consequences. Impact analysis is required before a CPR can be applied in clinical practice with confidence that its implementation will likely result in improved patient care.6,69,70,98 In the management of patients with LBP, the benefits of the clinical application of a well-developed diagnostic CPR may include reducing the need for unnecessary tests, identifying patients who are likely to benefit from referral to other services and/or further investigations, improving the efficiency of the clinical assessment, reducing the costs of care, enabling more timely initiation of treatment, and informing treatment decision making. Such benefits have been demonstrated or hypothesized for diagnostic CPRs developed for other musculoskeletal presentations.2,54,92,93,95,97,102 Previous systematic reviews of LBP CPRs have predominantly focused on prognostic tools designed to predict outcomes such as improvements in clinical status following treatment.7,66,67,73,91,101 To the best of our knowledge, only 1 previous review37 has included diagnostic forms of LBP CPRs; however, that study was limited to tools developed within physical therapy practice. Consequently, little is known about the current state of research on diagnostic LBP CPRs and their readiness to be applied in clinical practice. Therefore, the aim of the present review was to identify diagnostic forms of LBP CPRs and to appraise their readiness for application in clinical practice.

62 | february 2015 | volume 45 | number 2 | journal of orthopaedic & sports physical therapy

45-02 Haskins.indd 62

1/22/2015 10:44:23 AM

METHODS

Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Uni of Sydney Cumberland on November 24, 2016. For personal use only. No other uses without permission. Copyright © 2015 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

Literature Search

A

systematic search of the literature was conducted to identify derivation, validation, and impact analysis studies investigating diagnostic forms of CPRs relevant to the nonsurgical management of adults with LBP. The database search strategy (APPENDIX A, available online at www.jospt.org) incorporated search strings identified to have high sensitivity for predictionmodel studies in combination with disease-specific filters for back-related disorders.8,31,44,46 Components of this search strategy have been used in previous systematic reviews for prognostic CPRs.7,37,67,101 MEDLINE, Embase, the Cochrane Central Register of Controlled Trials, PyscINFO, CINAHL, AMED, and the Index to Chiropractic Literature were searched from their inception to July 2013. Identified records were downloaded into reference management software (EndNote X6.0.1; Thomson Reuters, New York, NY), and duplicates were removed. Hand searching and citation tracking were used as supplementary search strategies.

Study Selection A CPR was operationally defined as “a clinical tool that quantifies the individual contributions that various components of the history, physical examination, and basic laboratory results make towards the diagnosis, prognosis, or likely response to treatment in an individual patient.”69 The eligibility criteria used in this review are summarized in TABLE 1. Diagnostic CPRs were operationally defined as relating to the present status or classification of an individual, which included, but was not limited to, pathoanatomic diagnoses. Studies developing CPRs that function to inform predictions related to future outcomes or treatment effects were excluded from this review. Study eligibility was not restricted by year of publication, stage of CPR development, types of diagnostic predictor variables under consideration

TABLE 2

Japanese Society for Spine Surgery and Related Research Clinical Prediction Rule to Assist in the Identification of Lumbar Spinal Stenosis 50,55*

Variable/Status

Score

Age, y 70

2

Comorbidity of diabetes Present

0

Absent

1

Intermittent claudication (from history) Absent

0

Present

3

Exacerbation of symptom(s) when standing up (from history) Absent

0

Present

2

Symptom improvement when bending forward (from history) Absent

0

Present

3

Symptom(s) induced by having a patient bend forward (from physical examination) Present

–1

Absent

0

Symptom(s) induced by having a patient bend backward (from physical examination) Absent

0

Present

1

Peripheral artery circulation (from physical examination) Dorsalis pedis artery not easily palpable or ankle brachial index 30 min in duration Improvement in back pain with exercise but not with rest A  wakening because of back pain during the second half of the night only A  lternating buttock pain

*Presence of 2 or more variables is commonly used to denote positive status on this tool.

Statistical Analysis Interrater agreement for quality appraisal and each stage of study selection was calculated and reported as the absolute and chance-corrected (κ) degree of agreement. Two-by-two contingency tables were extracted, calculated, or approximated when reported study findings permitted. Sensitivity, specificity, likelihood ratios, and posterior probabilities and their associated uncertainty intervals were subsequently calculated from these data and reported for each study in instances where these were not directly conveyed in the original publication. The objective Bayesian method using Monte Carlo simulation was used to derive uncertainty intervals for posterior probabilities.36,71

Quality Appraisal The QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies-2) tool105 was applied by 2 independent reviewers, with consensus determining criterion status. A third reviewer provided the final decision in instances where consensus was not reached. Included studies were further appraised in regard to the methodological standards of CPR development specific to each phase of development. In the absence of a validated tool for this purpose, an updated version of a quality appraisal tool used in a previous systematic review on CPRs was applied (APPENDIX B, available online at www. jospt.org).37 The items selected for inclusion in this tool reflect well-recognized methodological standards for the development of CPRs.6,11,18,19,36,63,66,70,86,94,101

]

TABLE 4

• • • • •

“IBP According to Experts” Clinical Prediction Rule to Assist in the Identification of Inflammatory Back Pain 90*

A  ge at onset 52 y; (2) no patients with probpresence of leg pain; able osteoporotic (3) BMI ≤22 kg/m2; vertebral compres(4) does not exercise sion fracture regularly; (5) female sex

Count of predictors

Derivation n = 1448 patients with a Roman et al78 primary low back pain complaint presenting to tertiary care institution. Mean age, 55 y (calculated); 59% female; duration of symptoms not reported; 3% prevalence of dependent outcome

Spinal stenosis 3. To indicate the (1) bilateral symptoms; likelihood of the (2) leg pain more presence of lumbar than back pain; (3) spinal stenosis pain during walking/ standing; (4) pain relief on sitting; (5) age >48 y

Vertebral fracture …

Diagnosis of lumbar If 2 of 5 present: compression Sn, 0.95 (95% fracture or CI: 0.83, 0.99); wedge deforSp, 0.34 (95% mity by standard CI: 0.33, 0.34); radiographs –LR = 0.16 or CT scan (95% CI: 0.04. (dichotomous) 0.51) If 4 of 5 present: Sn, 0.37 (95% CI: 0.24, 0.51); Sp, 0.96 (95% CI: 0.95, 0.97); +LR = 9.6 (95% CI: 3.7, 14.9)

Scores ≥7, probability of fracture ≥43% Scores ≤2, probability of fracture 30 Count of min; (2) improvement predictors in back pain with exercise but not with rest; (3) awakening because of back pain during the second half of the night only; (4) alternating buttock pain

Diagnosis of For 2 or more Posttest probability for inflammapredictors pres2 or more predictors tory back pain by ent: Sn, 0.70; present: not reported, rheumatologist Sp, 0.81 but approximated (dichotomous) LRs not reported, from study data to but approximatbe 88% ed to be: +LR 95% CrI approximated = 3.8 (95% CI: from study data to be 2.8, 5.0); –LR = 84.1%, 91.0% 0.37 (95% CI: 0.31, 0.43) For 2 or more Posttest probability for predictors pres2 or more predictors ent: Sn, 0.92; present: not reported, Sp, 0.67 but calculated from LRs not reported, study data to be 75% but calculated 95% CrI calculated from to be +LR = 2.8 study data to be (95% CI: 1.2, 49.7%, 89.9% 6.3); –LR = 0.12 (95% CI: 0.02, 0.79) Table continues on page 68.

journal of orthopaedic & sports physical therapy | volume 45 | number 2 | february 2015 | 67

45-02 Haskins.indd 67

1/22/2015 10:44:25 AM

[

]

Characteristics of the Diagnostic Low Back Pain Clinical Prediction Rules Included in This Review (continued)

TABLE 5 Diagnosis Supported/ CPR/Function Predictors Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Uni of Sydney Cumberland on November 24, 2016. For personal use only. No other uses without permission. Copyright © 2015 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

research report

Format

Study

Patients

Reference Standard Accuracy

Posttest Probability

Count of predictors

Derivation Sieper et al90

n = 20 patients selected Diagnosis of For 4 or more by rheumatologists with inflammapredictors suspected axial SpA. tory back pain by present: Sn, Mean  SD age, 41  11 rheumatologist 0.77; Sp, 0.92 y; 60% female; duration with expertise LRs not reported, of symptoms, 7  6 y in AS/SpA but calculated (dichotomous) to be +LR = 9.2 (95% CI: 3.6, 23.9); –LR = 0.25 (95% CI: 0.16, 0.40)

Count of predictors

Validation Sieper et al90

n = 648 patients with chronic back pain of unknown origin that began before 45 y of age presenting to a rheumatologist. Mean age, 34 y; 55% female; mean duration of symptoms, 7.3 y; approximated prevalence of dependent outcome, 66%

Diagnosis of For 4 or more Posttest probability for inflammapredictors 4 or more predictors tory back pain by present: Sn, present: not reported, rheumatologist 0.80; Sp, 0.72 but approximated (dichotomous) LRs not reported, from study data to but approximatbe 85% ed to be +LR 95% CrI approximated = 2.9 (95% CI: from study data to be 2.3, 3.6); –LR = 81.0%, 88.0% 0.28 (95% CI: 0.23, 0.35)

Count of predictors

Derivation Braun et al13

n = 322 patients with low back pain >2 mo presenting to orthopaedic surgeons, most with ≥1 inflammatory back pain symptom. Mean  SD age, 36  8 y; 51% female; duration of symptoms, 44  38 mo; 35% prevalence of dependent outcome

Diagnosis of axial SpA as determined by a rheumatologist (dichotomous)

Inflammatory back pain 7 (IBP according to (1) age at onset 65 y; (b) pain not exacerbated by coughing; (c) not worsened by hyperextension; (d) not worsened by forward flexion; (e) not worsened when rising from flexion; (f) not worsened by extension-rotation; (g) well relieved by recumbency

For both predictors Posttest probability present: Sn, for both predictors 0.92; Sp, 0.80 present: not reported, LRs not reported, but approximated but approximatfrom study data to ed to be +LR be 67% = 4.5 (95% CI: 95% CrI approximated 2.2, 9.3); –LR = from study data to be 0.10 (95% CI: 42.2%, 83.3% 0.02, 0.64)

Radicular low back pain 10 (StEP). Differentiation of radicular and axial low back pain

(1) radicular pain in Score chart the SLR test; (2) abnormal response to cold temperature; (3) abnormal response to pinprick; (4) abnormal response to blunt pressure; (5) decreased response to vibration; (6) dysesthesia; (7) temporal summation; (8) burning or cold quality of pain; (9) abnormal response to brush movement; (10) ongoing pain; (11) skin changes

Derivation n = 137 (described as Radicular low back Score ≥4 points: Scholz et al85 validation set in study) pain as deterSn, 0.92 (95% patients with chronic mined by team CI: 0.83, 0.97); low back pain referred to consisting of Sp, 0.97 (95% neurosurgical outpatient rheumatologist, CI: 0.89, 1.0) triage clinic. Demographneurosurgeon, LRs not reported, ics for radicular low back and physiotherabut calculated pain group: median age, pist (dichototo be +LR = 45 y (range, 20-82 y); mous) 28.5 (95% CI: 55% female; median, 7.3, 111.7); –LR 1 y (range, 0.3-34 y) of = 0.08 (95% symptoms. For axial low CI: 0.04, 0.18) back pain group: median age, 55 y (range, 24-78 y); 56% female; median duration of symptoms, 5 y (range, 0.3-46 y); 55% prevalence of dependent outcome

(1) lack of hypomobility during intervertebral testing; (2) lumbar flexion ROM >53°

Derivation Fritz et al30

For scores ≥4 points: 97% (90%, 100%) 95% CrI for scores ≥4 points calculated from study data to be 90.3%, 99.1%

Radiographic instability 11. Identification of patients with probable radiographic instability

Count of predictors

n = 49 patients with low 2 segments with back pain referred for either rotational imaging on suspicion or translational of instability. Mean  instability or 1 SD age, 39  11 y; 57% segment with female; median duration both translaof symptoms, 78 d tional and rota(variability not reported); tional instability 57% prevalence of on radiographic dependent outcome assessment (dichotomous)

For both predictors For both predictors present: Sn, present: 93% 0.29 (95% CI: 95% CrI for both 0.13, 0.46); Sp, predictors present: 0.98 (95% CI: calculated from study 0.91, 1.0); +LR data to be 66.6%, = 12.8 (95% CI: 99.7% 0.79, 211.6); – LR = 0.72 (95% CI: 0.55, 0.94) Table continues on page 70.

journal of orthopaedic & sports physical therapy | volume 45 | number 2 | february 2015 | 69

45-02 Haskins.indd 69

1/22/2015 10:44:26 AM

[

]

Characteristics of the Diagnostic Low Back Pain Clinical Prediction Rules Included in This Review (continued)

TABLE 5 Diagnosis Supported/ CPR/Function Predictors Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Uni of Sydney Cumberland on November 24, 2016. For personal use only. No other uses without permission. Copyright © 2015 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.

research report

Format

Study

Patients

Reference Standard Accuracy

Count of predictors

Derivation Gregg et al33

n = 82 patients with low Diagnosis of active back pain with suspected spondylolysis spondylolysis referred by SPECT bone for SPECT bone scan; scan (dichoto66% under 20 y of age; mous) 48% female, 52% male; duration of symptoms, 3 mo attending outpatient rehabilitation. Mean  SD age, 44  11 y; 63% female; median duration of symptoms, 84 mo (IQR, 24-180 mo); 53% prevalence of dependent outcome

Posttest Probability

Spondylolysis 12. Identification of patients with probable active spondylolysis

(1) male sex; (2) age ≤20 y

For both predictors Posttest probability for present: Sn, both predictors pres0.62; Sp, 0.84; ent: not reported, but +LR = 3.66; calculated from study –LR = 0.27 data to be 64% 95% CrI calculated from study data to be 43.8%, 79.5%

Psychological disturbance 13. Identification of patients with chronic low back pain who might benefit from additional psychological assessment

(1) number of Waddell Score chart signs (0-8); (2) pain drawing score (0-100); (3) no directional preference; (4) daily use of pain medication for chronic low back pain

Relevant psychological disturbance as determined by a psychologist (dichotomous)



Score ranges matched to posttest probability ranges;

Diagnostic clinical prediction rules for specific subtypes of low back pain: a systematic review.

Systematic review...
1MB Sizes 0 Downloads 9 Views