Advertisement

Prognostic Models in Adults Undergoing Physical Therapy for Rotator Cuff Disorders: Systematic Review

Cordula Braun, Nigel C. Hanchard, Alan M. Batterham, Helen H. Handoll, Andreas Betthäuser
DOI: 10.2522/ptj.20150475 Published 1 July 2016

Abstract

Background Rotator cuff–related disorders represent the largest subgroup of shoulder complaints. Despite the availability of various conservative and surgical treatment options, the precise indications for these options remain unclear.

Purpose The purpose of this systematic review was to synthesize the available research on prognostic models for predicting outcomes in adults undergoing physical therapy for painful rotator cuff disorders.

Data Sources The MEDLINE, EMBASE, CINAHL, Cochrane CENTRAL, and PEDro databases and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) up to October 2015 were searched.

Study Selection The review included primary studies exploring prognostic models in adults undergoing physical therapy, with or without other conservative measures, for painful rotator cuff disorders. Primary outcomes were pain, disability, and adverse events. Inclusion was limited to prospective investigations of prognostic factors elicited at the baseline assessment. Study selection was independently performed by 2 reviewers.

Data Extraction A pilot-tested form was used to extract data on key aspects of study design, characteristics, analyses, and results. Risk of bias and applicability were independently assessed by 2 reviewers using the Prediction Study Risk of Bias Assessment tool (PROBAST).

Data Synthesis Five studies were included in the review. These studies were extremely heterogeneous in many aspects of design, conduct, and analysis. The findings were analyzed narratively.

Limitations All included studies were rated as at high risk of bias, and none of the resulting prognostic models was found to be usable in clinical practice.

Conclusions There are no prognostic models ready to inform clinical practice in the context of the review question, highlighting the need for further research on prognostic models for predicting outcomes in adults who undergo physical therapy for painful rotator cuff disorders. The design and conduct of future studies should be receptive to developing methods.

Painful shoulder complaints are among the most common musculoskeletal disorders in medical and physical therapist practice.1 These complaints may become persistent, potentially leading to increased use of health care resources and prolonged sick leave and placing a cost burden on the individual and society.2,3 Most shoulder complaints (29%–85%) involve the subacromial-subdeltoid bursa and rotator cuff.4,5 The pathology is diverse, reflecting a degenerative continuum from tendinopathy to partial- or full-thickness tears.6 Rotator cuff tears, in particular, have a reported prevalence of over 40% in symptomatic shoulder pain populations7 and are strongly correlated with age.8 Clinical features of rotator cuff disorders may include pain, abnormalities on tests of rotator cuff function and integrity,9 and significantly impaired shoulder function and health-related quality of life.10,11 Although diagnosis of rotator cuff disorders is based on clinical signs and symptoms,9 verification of a rotator cuff tear requires diagnostic imaging (eg, ultrasonography, magnetic resonance imaging).12

Initial treatment of rotator cuff disorders usually involves medical care (eg, oral medication, corticosteroid injections) and physical therapy (eg, exercises, manual therapy). Current guidelines advise conservative treatment as the first-line treatment, with surgery mainly reserved for nonresponders.13–15 Direct comparisons of conservative versus surgical treatment16–19 have not shown clinically relevant differences between groups. Nonetheless, the rates of surgical intervention for rotator cuff disease have considerably increased in many countries.20,21 Unnecessary surgery is undesirable, as is ineffective conservative treatment. Patients and health care providers alike would benefit if likely responders and, by corollary, nonresponders to conservative interventions could be identified at the commencement of the care pathway. This early identification would avoid unnecessary suffering, reduce uncertainty and anxiety, limit exposure to the risks of surgery, and conserve limited resources. “Understanding which patients [with rotator cuff tears] do best with nonoperative treatment” has been rated a top priority research issue.22

The importance of predicting which patients will respond to particular treatments is increasingly recognized and has stimulated interest in prognosis and prognosis research.23 There has been a corresponding development in prognosis research methodology.24–26 Prognosis research aims to predict clinical outcomes in individual patients.25 One aspect of prognosis research involves single factors, which, in the context of painful rotator cuff disorders, would typically be demographic or clinical. However, single factors are unlikely to predict outcomes satisfactorily. Multivariable prognostic models are better placed to do so because they account for real-life clinical complexities.27,28 An illustration of a multivariable model is the Nottingham Prognostic Index (NPI), which is used to predict survival of women diagnosed with primary breast cancer by the following formula: NPI = [0.2 × tumor diameter (cm)] + lymph node stage + tumor grade.29 Scores are interpreted by reference to a table.

Prognostic modeling encompasses 3 key phases: development (including internal validation [ie, determining the model's replicability using data from the primary sample]), external validation (determining the model's generalizability using data from independent samples), and investigation of clinical impact (a model's effectiveness and cost-effectiveness in improving outcomes).28,30 External validation is a crucial step before a model can be considered usable in clinical practice.28

The objective of this review was to synthesize the available research on prognostic models for predicting outcomes in adults who undergo physical therapy for painful rotator cuff disorders. We aimed to provide a resource to facilitate clinical decision making and to identify any research gaps. To our knowledge, this is the first systematic review to synthesize the available evidence on this topic.

Method

Overall Approach

We based our methods on the recent recommendations of the PROGRESS (PROGnosis RESearch Strategy) partnership25 and, complementarily, the Cochrane Prognosis Methods Group.26 We used PROGRESS terminology where possible.27 This review is based on an a priori protocol, registered in PROSPERO, the International Prospective Register of Systematic Reviews31 (registration number: CRD42014008973), and available at http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42014008973. Differences between protocol and review are specified within the supplementary material (eTab. 1).

Criteria for Considering Studies for Inclusion

Types of studies.

We included primary studies exploring prognostic models for predicting outcomes in adults undergoing physical therapy, with or without other conservative measures, for painful rotator cuff disorders. Inclusion encompassed any of the 3 phases of prognostic research. We considered any prospective longitudinal research designs. There was no language restriction on searches. Only reports written in English were included, but we planned to document relevant studies reported in other languages.

Participants.

This review addressed adults (age ≥18 years) diagnosed with painful rotator cuff disorders, at any stage, which was unrelated to substantial trauma (eg, dislocation). We placed no restriction on how these disorders were diagnosed. We also included studies whose inclusion criteria were symptoms or mechanisms consistent with rotator cuff disorders (eg, subacromial pain, subacromial impingement, shoulder impingement). Studies in which 85% or more of the participants satisfied our criteria were included. We did not actively seek studies focused on subacromial-subdeltoid bursitis, although, due to its intimate relationship with the rotator cuff, incidental involvement of this bursa may well occur in our population of interest. There was no restriction on the duration or severity of symptoms at baseline, or on the care setting.

We excluded studies focusing on people who were pain-free or had trauma-related conditions and studies on calcific tendinitis or disorders of the long head of the biceps. We anticipated that, in some studies, there would be insufficient characterization of participants (eg, that other potential causes of shoulder pain might not be considered). In these cases, we erred on the side of inclusivity.

Interventions.

We included studies evaluating physical therapy, of any duration or frequency and with or without other conservative measures, as part of a nonsurgical care pathway. Physical therapy had to involve therapeutic exercises or manual techniques, as these are considered the core interventions,32 but could include adjunctive treatments (eg, acupuncture, electrotherapy, corticosteroid injections, osteopathic musculoskeletal interventions, thermotherapy). Studies comparing physical therapy and a non–physical therapy control group were considered only if there was separate prognostic modeling for physical therapy.

Prognostic factors.

For simplicity, we applied the term “prognostic factor” to any factor under investigation, regardless of whether it was (or had previously been) found to have prognostic properties. We required these factors to be elicited at the baseline assessment.

Outcomes.

Primary outcomes were: pain, shoulder disability on a validated patient-reported outcome measure (eg, Oxford Shoulder Score), and adverse events (eg, exacerbations of symptoms). Secondary outcomes were: health-related quality of life (eg, 36-Item Short-Form Health Survey [SF-36]), sick leave, the patient's global perception of change (GPC), imaging determination of structural progression of tear, and the patient's decision to undergo surgery. To be included, a study had to present a prognostic model in relation to at least one of these outcomes.

Types of analysis.

Studies had to evaluate prognostic models of multiple factors, but no restriction was placed on the phase of research or on the type of multivariable analysis. Furthermore, the models had to be presented in full in the study report or provided on request by the corresponding authors.

Data Sources and Searches

Electronic searches.

Building on the experience of previous searches for a prognostic study (2011–2012, report in preparation) and 2 systematic reviews of interventions in this field,32,33 we developed a broad strategy including only search terms relating to the population and interventions. For MEDLINE, we used a slightly amended version of a filter developed for prognosis research34 (see eTab. 2 for the full search strategy).

We searched the following electronic databases from inception: MEDLINE (EBSCO), EMBASE (Ovid), Cochrane CENTRAL (Ovid), CINAHL (EBSCO), PEDro, and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP). The formal database search was initially run on May 16, 2014 (ICTRP was searched on August 14) and updated to October 19, 2015. One author (C.B.) conducted the searches. We followed up “related articles” suggestions for all relevant studies.

Searching other sources.

We supplemented the electronic searches by hand searching the reference lists of all relevant studies and existing prognosis systematic reviews on shoulder pain. We further matched the compilation of eligible studies with the results from our previous searches.

Study Selection

Study selection was independently performed by 2 authors (C.B. and N.C.H. or H.H.H.). In case of disagreement, consensus was sought through discussion or involvement of a third person (A.M.B. or H.H.H.).

Data Extraction and Quality Assessment

Data extraction and management.

We used a purpose-designed and pilot-tested form to extract data on key aspects of study design, characteristics, analyses, and results. For developmental studies, we extracted only one model per study: either the reportedly final model or the most complete model including the main effects for all prognostic factors. We extracted key statistics of the models and of model performance as reported by the studies. Extraction of summary statistics of predictive performance, where possible, included the standard error of the estimate (SEE) for studies with continuous outcomes and likelihood ratios or area under the curve (C statistics) for studies with binary outcomes. We also reported any further measures of model performance (eg, of the model's discriminative ability) and validation (internal or external). Two authors (C.B., N.C.H.) independently extracted the data. We did not impute missing data. We limited author contact to the clarification of issues related to study eligibility.

Assessment of risk of bias and applicability.

To assess risk of bias and applicability, we used the latest available version of the Prediction Study Risk of Bias Assessment Tool (PROBAST),35 which, at the time of writing, was in the late stages of development but unpublished (Robert Wolff; personal communication; September 16, 2014). The PROBAST is designed to assess risk of bias and applicability of primary studies evaluating (developing or validating) prognostic models. It is domain-based, with a similar structure to that of the QUADAS-2.36 It has 5 key domains: participant selection, predictors (ie, prognostic factors), outcome, sample size and participant flow, and analysis. Each domain comprises a set of “signaling questions” to facilitate judgments about risk of bias: low, high, or unclear. Additionally, the first 3 domains are assessed for concerns (low, high, or unclear) about the applicability of the study's design and characteristics to the review question. A summative judgment across all domains leads to an overall rating of low, high, or unclear risk of bias or concern about applicability. Lastly, the usability of the model is rated as “yes” or “no.” For this item, we considered whether the model was ready for use in the intended context and target population, in view of the phase of research, the detail with which the model was presented, and the risk of bias. Risk of bias and applicability assessment was independently performed by 2 authors (C.B., N.C.H.). In case of disagreement, consensus was sought through discussion or involvement of a third person (A.M.B. or H.H.H.).

Data Synthesis and Analysis

All included studies were tabulated and narratively synthesized. In the absence of sufficient good-quality, comparable, and externally validated studies, we did not undertake quantitative data synthesis.

Results

Search and Selection Process

The complete process is outlined in the Figure. The titles and abstracts of 5,889 results overall were screened. Fifty-four full-text articles were obtained and considered for inclusion, 6 of which were identified from previous prognosis systematic reviews,37,38 5 from our previous searches, and 1 from personal communication (Thilo Kromer; September 22, 2014). We included 5 studies39–43 and excluded 49 studies (see eTab. 3 for further details). The most frequent reason for exclusion was a lack of multivariable prognostic modeling. We identified (by protocol or registry entries) 8 clearly or potentially relevant ongoing studies (see eTab. 4 and Figure). We obtained unpublished full multivariable model data relating to the trial of Hallgren et al.40

Figure.
Figure.

Search and selection flow diagram (adapted from Liberati et al59).

Included Studies

Study characteristics.

Key characteristics of the 5 studies are presented in Table 1. The studies were published between 2005 and 2014. All studies appeared to have been conducted in outpatient settings. All were cohort studies, but in 2 studies, the cohort was derived from pooled data from a randomized controlled trial (RCT).39,40 None of the studies were prospectively registered; however, the intention for a prognostic investigation was mentioned in the published protocol for the study by Kromer et al.44 Four studies concerned model development, and the fifth study42 was reported as a validation study.

View this table:
Table 1.

Characteristics of Included Studies (Alphabetical Order)a

Four studies39–41,43 investigated mixed populations with impingement-related shoulder pain. One of these studies41 excluded full-thickness tears. One study42 investigated a rotator cuff tear population without differentiating between partial- and full-thickness tears. Initial sample sizes ranged from 3341 to 102,40,43 with the number of outcome events (number of patients in whom the prognosticated event occurred) ranging from 2341 to 89.43 Although varying in duration, content, and dosage, physical therapy was provided to all study participants; steroid injections were provided to all participants of one study40 and were optional in another study.43

The number of initially considered prognostic factors was unclear in 3 studies,39–41 but, based on the presented data, appeared to range from 8 prognostic factors43 to presumably more than 60.41 Prognostic factors mainly involved demographics and clinical characteristics, such as symptoms or diagnostic imaging findings. One study39 investigated psychosocial factors. None of the studies provided a full and unambiguous rationale for all initially considered factors. Although, in some cases, the authors referred to previous prognosis research, the approaches to the literature appeared nonsystematic. Kromer et al39 presented some focused, literature-based justification for 2 of the factors modeled: fear-avoidance beliefs and catastrophizing. Apart from these 2 exceptions, prognostic factors were not systematically derived from the literature.39–43

Each study used different outcome measures, but all studies included patient-reported outcome measures; the outcomes used for this review are presented in Table 1. Follow-up ranged from 6 weeks41 to 12 months.39,40,42,43

The methods for selecting prognostic factors for multivariable analysis, where specified, varied (eTab. 5); 2 studies39,41 explicitly reported using some automated statistical method (eg, analysis of univariable correlations between the prognostic factors and the outcome).

Approaches to multivariable modeling also varied. An automated statistical process (eg, stepwise regression) was used in 3 studies.39,41,43 The nominal validation study by Merolla et al42 was severely flawed by inappropriate statistical analysis.

Risk of bias and applicability.

Table 2 presents the summary of our PROBAST assessment. Overall, all studies were rated as at high risk of bias, mainly due to issues within domains 3 to 5 (outcome, sample size and flow, and analysis). Ratings were affected by numerous issues, namely: inclusion of prognostic factors in the outcome definition39,42,43; unclear or lack of blinding of outcome determination to prognostic factor information41,42; an unreasonable number (>5) of prognostic factors in relation to the number of outcome events (which we assessed in relation to the number of factors included in the reported final model or, where this number was not specified, the most complete model including main effects for all prognostic factors)40,42; unclear handling of missing data39–43; use of univariable analyses to select prognostic factors39,41; unclear40 or unspecified42 modeling methods; and failure to consider overfitting of data, complexities in the data, evaluation of performance measures, or nonlinear relationships.39–43

View this table:
Table 2.

Prediction Study Risk of Bias Assessment Tool (PROBAST) (Risk of Bias and Applicability) Ratingsa

Overall, concerns about applicability mainly related to domain 2 (predictors) and were rated as low for 2 studies,39,40 unclear for 1 study,43 and high for 2 studies (for concerns related to the definition and assessment of prognostic factors).41,42 We rated all models as not usable in clinical practice (Tab. 2). Both risk of bias and applicability ratings were affected by inadequate reporting, which was a primary reason for “unclear” domain ratings.

Results of included studies.

Heterogeneity of clinical characteristics, prognostic factors, and methods, including the statistical approaches to multivariable modeling, precluded the statistical synthesis of the 4 development studies and limited the narrative synthesis of all 5 studies. Considering the studies' heterogeneity and poor performance against the PROBAST criteria, we limited the presentation of data within our review to a table of key study characteristics without results (Tab. 1). For a more detailed table of the characteristics, including results, see eTable 5.

The presented models differed greatly in various aspects, including the number and composition of prognostic factors, and in the presented statistics (eTab. 5). Only Hung et al41 provided a prognostic index (eTab. 5).

Conflicts of Interest

Conflicts of interest were explicitly addressed only in 2 studies,40,43 which stated that there were none.

Discussion

Summary of Main Results

This systematic review included 5 studies with a total of 387 patients that aimed to either develop39–41,43 or validate42 prognostic models for predicting outcomes in adults who undergo physical therapy, with or without other conservative measures, for painful rotator cuff disorders.

The studies were heterogeneous in terms of the populations, phases of research, prognostic factors studied, statistical approaches used, and results reported. These considerations ruled out meaningful statistical synthesis and imposed major limitations on narrative synthesis. Moreover, all of the studies were at high risk of bias, and most raised “unclear” or “high” concerns about applicability. None of the models were ready for use in practice.

Overall Completeness, Applicability, and Usability of the Evidence

The study populations were broadly relevant to the review question. Four studies39–41,43 investigated populations with impingement-related shoulder pain, implicitly including rotator cuff tears of varying completeness, except for the study by Hung et al,41 who excluded full-thickness tears. Merolla et al42 exclusively studied rotator cuff tears, although it is unclear whether they incorporated partial-thickness tears in this definition. However, applicability was compromised by unclear eligibility criteria in some studies, pertaining, for example, to frozen shoulder41 or rotator cuff tears.39,40,43 Also, in 2 studies, the patient populations were selected by dint of their agreement to participate in an RCT,39,40 which may have reduced external validity.

The physical therapy intervention was insufficiently described to allow a judgment in the study by Taheriazam et al.43 However, in the intervention group of Hallgren et al40 and in the other 3 studies,39,41,42 the physical therapy intervention was generally consistent with standard practice.45,46

Less uniform was the selection of predictors, which was generally unjustified and diverse. In one case,41 prediction required measurement using specialized equipment (the FASTRAK motion analysis system, Polhemus Inc, Colchester, Vermont) that would not be available in most clinical settings. Replicability and applicability of the models are likely to be reduced by the questionable clinimetric properties of some prognostic factor measurements, such as posterior shoulder tightness in the study by Hung et al41 and the application of arbitrary cutoff points for categorizing continuous prognostic factors.

Some of our prespecified outcomes were reported in some studies, including pain,42 shoulder disability,39,42,43 and global perceived change.41 Hallgren et al40 reported the decision to undergo surgery. The remaining outcomes of interest for this review, including adverse events, health-related quality of life, sick leave, and structural progression of tears, were either not reported or, in one case,42 reported too unclearly for extraction.

None of the 4 development studies39–41,43 reported any form of internal model validation, and none of these studies were followed by an external validation, even though 5 and 10 years had elapsed since the studies by Hung et al41 and Taheriazam et al,43 respectively. Lack of appropriate validation of prognostic models is a widely observed issue.47 There is good empirical evidence that models perform substantially less well in external (ie, independent) samples and that performance in external samples is more representative of clinical performance,28,48 so these factors present a major obstacle to usability. The fifth study (Merolla et al42), although reportedly a validation study, was seriously flawed in both concept and execution. Ultimately, none of the studies have been assessed for clinical impact, and, consequently, none of the models presented in the included studies is usable in clinical practice.

Quality of the Evidence

We evaluated risk of bias in 5 domains: participant selection, predictors, outcome, sample size and flow, and analysis. Our judgment of risk of bias was affected by a number of methodological issues (see Results section). Most of the identified deficiencies have been addressed extensively in the literature; several, including, in particular, those relating to the number of prognostic factors in relation to the number of outcome events and use of univariable analyses to select prognostic factors, have been shown to result in invalid and unreliable models.49 Similarly, the use of statistical methods, such as stepwise regression, to select factors within the multivariable analysis has been criticized.49,50 These findings suggest that the presented models are highly unlikely to produce valid and reliable predictions. Moreover, deficiencies such as unclear handling of missing data and the failure to consider overfitting of data, complexities in the data, evaluation of performance measures, or nonlinear relationships seriously hamper the judgment of the quality of the data and the models' performance. The single “validation” study, by Merolla et al,42 was at high risk of bias in most domains.

An issue warranting special emphasis is the inclusion of prognostic factors in the outcome definition (ie, the problem of incorporation bias through mathematical coupling), as it represents a conflict between risk of bias and applicability. The literature on incorporation bias primarily relates to diagnostic research. In that context, it relates to the interaction between index and reference tests.51 Mathematical coupling, which inherently occurs “when one variable directly or indirectly contains the whole or part of another,”52(p444) may either erroneously purport a relationship between the prognostic factors and the outcome or overestimate an existing relationship, thus inflating estimates of predictive performance. The conflict with applicability arises specifically because baseline and endpoint evaluation of a given outcome measure is standard clinical practice. This approach particularly applies to the increased use of patient-reported outcome measures in clinical practice and research.53 Moreover, in the present context, patient-reported outcome measures are among very few prognostic factors that have a basis in evidence.37,38 In our review, this conflict was encountered in 2 studies,39,43 which were both downgraded for risk of bias in the outcome domain. The described problem may be accommodated in the study design (eg, by including a no-treatment control group as a point of reference) or addressed at the analysis stage, but should not be overlooked.

Potential Biases in the Review Process

We sought to minimize bias in the review process by developing an a priori protocol that was registered with PROSPERO. In addition, the full protocol was lodged, a priori, with the chair of the Research Governance and Ethics Committee of the School of Health and Social Care at Teesside University. We recorded any deviations from the protocol (eTab. 1).

Our searches were comprehensive and included several supplementary sources, as well as the thorough inspection of all search results. The known difficulty of identifying prognosis research34,54 is reflected by the <0.1% yield of included studies from our initial results (Figure). Problems include the lack of appropriate indexing functions in the electronic databases and of current validated search filters. We identified several search filters for prognosis research, 3 of which are supplied here34,55,56; however, we had concerns about the currency of all search filters but one (ie, MEDLINE),34 which was purposely designed to identify prognostic model studies for systematic reviews. Applying this filter (amended by “prognos*”) significantly decreased the number of results in MEDLINE, but nonetheless, in contrast to all other databases searched, retrieved all 5 studies that were included in this review. This finding suggests that this filter performs well. Identification of relevant studies also was hampered by uninformative titles and abstracts, and inconsistent terminology compounded these difficulties, as has been noted by other authors.27,28 Although we restricted inclusion to reports in English, we did not impose a language restriction on our searches and did not identify any non-English, but clearly relevant, studies.

Systematic reviewing of prognostic modeling studies is an evolving field, and the methodology is a work in progress. Nonetheless, in evaluating the studies, we referred to the latest recommendations of the PROGRESS partnership25 and, after pilot testing earlier versions, evaluated risk of bias and applicability using a near-definitive but unpublished version of PROBAST (Robert Wolff; personal communication; September 16, 2014). The use of PROBAST was especially appealing to us because it is the first tool to specifically address risk of bias and applicability in prognostic model studies.

Agreements and Disagreements With Other Studies or Reviews

To our knowledge, this is the first systematic review to synthesize the evidence on primary prognostic model research in adults with rotator cuff disorders who are undergoing conservative treatment with physical therapy. We identified 2 other prognostic systematic reviews addressing shoulder pain,37,38 but both aimed to synthesize evidence on individual prognostic factors rather than on prognostic models and have minimal overlap with our own review, which has a single study41 in common with the review by Chester et al37 and none with the review by Kuijpers et al.38 Of the 2 reviews, Chester et al,37 like us, limited inclusion to studies investigating response to conservative treatment with physical therapy, whereas Kuijpers et al38 studied overall prognosis. Both reviews addressed shoulder pain in general and did not provide any subgroup analyses to allow for inferences about rotator cuff disorders. Thus, although evidence was found supporting a limited number of emerging factors, including symptom duration, baseline function or disability,37,38 pain, and age,38 the transferability of these findings to the population of interest in our review is unclear.

Implications for Practice

There is no prognostic model ready to inform clinical practice on the prognosis of outcomes in adults who undergo physical therapy, with or without other conservative measures, for painful rotator cuff disorders.

Implications for Research

The complexity of prognostic modeling demands high levels of methodological expertise and clinical judgment, but particularly calls for the involvement, from the outset, of a statistician with expertise in the field. The composition of primary (and secondary) research teams, therefore, should reflect this need. Researchers should be receptive to developing methods that may improve the validity and reliability of prognostic models. Crucially, more attention should be paid to model validation, and ultimately, to the assessment of clinical impact.

The PROBAST,35 once publicly available, should facilitate the assessment of risk of bias and applicability in future systematic reviews of prognostic model studies. Furthermore, both methods and reporting will benefit from adherence to the recommendations set out in the recent TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) statement.57 Further guidance for systematic reviews of prognostic model studies is now available through the CHARMS (Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies) checklist.58

Footnotes

  • All authors provided concept/idea/research design and consultation (including review of manuscript before submission). Ms Braun, Dr Hanchard, Professor Batterham, and Dr Handoll provided writing and data analysis. Ms Braun, Dr Hanchard, Dr Handoll, and Dr Betthäuser provided data collection. Ms Braun provided project management.

  • The authors acknowledge the assistance of Robert Wolff, Kleijnen Systematic Reviews Ltd, York, United Kingdom. Dr Wolff shared with the authors the prepublished versions of PROBAST, which they pilot tested in the development of this review, and provided support on their use. The authors also acknowledge the support from Iain Baird and Julie Hogg, Library and Information Services, Teesside University, Middlesbrough, United Kingdom, on the development of the search strategy and Hanna Björnsson Hallgren for providing unpublished analysis data.

  • All authors are contributors to an unfunded primary prognostic modeling study of rotator cuff disease treated by physical therapy (registration no. DRKS00004462). The systematic review also was not funded.

  • Systematic review registration number (PROSPERO): CRD42014008973.

  • Received August 20, 2015.
  • Accepted November 22, 2015.

References

  1. ↵
    1. Kooijman M,
    2. Swinkels I,
    3. van Dijk C,
    4. et al
    . Patients with shoulder syndromes in general and physiotherapy practice: an observational study. BMC Musculoskelet Disord. 2013;14:128.
    OpenUrlCrossRefPubMed
  2. ↵
    1. Paloneva J,
    2. Koskela S,
    3. Kautiainen H,
    4. et al
    . Consumption of medical resources and outcome of shoulder disorders in primary health care consulters. BMC Musculoskelet Disord. 2013;14:348.
    OpenUrlCrossRefPubMed
  3. ↵
    1. Virta L,
    2. Joranger P,
    3. Brox JI,
    4. Eriksson R
    . Costs of shoulder pain and resource use in primary health care: a cost-of-illness study in Sweden. BMC Musculoskelet Disord. 2012;13:17.
    OpenUrlCrossRefPubMed
  4. ↵
    1. Östör AJK,
    2. Richards CA,
    3. Prevost AT,
    4. et al
    . Diagnosis and relation to general health of shoulder disorders presenting to primary care. Rheumatology (Oxford). 2005;44:800–580.
    OpenUrlAbstract/FREE Full Text
  5. ↵
    1. Van der Windt DA,
    2. Koes BW,
    3. de Jong BA,
    4. Bouter M
    . Shoulder disorders in general practice: incidence, patient characteristics, and management. Ann Rheum Dis. 1995;54:959–964.
    OpenUrlAbstract/FREE Full Text
  6. ↵
    1. Cook JL,
    2. Purdam CR
    . Is tendon pathology a continuum? A pathology model to explain the clinical presentation of load-induced tendinopathy. Br J Sports Med. 2009;43:409–416.
    OpenUrlAbstract/FREE Full Text
  7. ↵
    1. Reilly P,
    2. Macleod I,
    3. Macfarlane R,
    4. et al
    . Dead men and radiologists don't lie: a review of cadaveric and radiological studies of rotator cuff tear prevalence. Ann R Coll Surg Engl. 2006;88:116–121.
    OpenUrlCrossRefPubMedWeb of Science
  8. ↵
    1. Beaudreuil J,
    2. Bardin T,
    3. Orcel P,
    4. Goutallier D
    . Natural history or outcome with conservative treatment of degenerative rotator cuff tears. Joint Bone Spine. 2007;74:527–529.
    OpenUrlCrossRefPubMedWeb of Science
  9. ↵
    1. Hanchard NC,
    2. Lenza M,
    3. Handoll HH,
    4. Takwoingi Y
    . Physical tests for shoulder impingements and local lesions of bursa, tendon or labrum that may accompany impingement. Cochrane Database Syst Rev. 2013;4:CD007427.
    OpenUrlPubMed
  10. ↵
    1. Minns Lowe CJ,
    2. Moser J,
    3. Barker K
    . Living with a symptomatic rotator cuff tear “bad days, bad nights”: a qualitative study. BMC Musculoskelet Disord. 2014;15:228.
    OpenUrlCrossRefPubMed
  11. ↵
    1. Piitulainen K,
    2. Ylinen J,
    3. Kautiainen H,
    4. Häkkinen A
    . The relationship between functional disability and health-related quality of life in patients with a rotator cuff tear. Disabil Rehabil. 2012;34:2071–2075.
    OpenUrlCrossRefPubMed
  12. ↵
    1. Lenza M,
    2. Buchbinder R,
    3. Takwoingi Y,
    4. et al
    . Magnetic resonance imaging, magnetic resonance arthrography and ultrasonography for assessing rotator cuff tears in people with shoulder pain for whom surgery is being considered. Cochrane Database Syst Rev. 2013;9:CD009020.
    OpenUrlPubMed
  13. ↵
    American Academy of Orthopaedic Surgeons (AAOS). Optimizing the management of rotator cuff problems: guidelines and evidence report. Available at: http://www.aaos.org/research/guidelines/RCP_guideline.pdf. December 4, 2010. Accessed June 21, 2015.
  14. ↵
    1. Beaudreuil J,
    2. Dhénain M,
    3. Coudane H,
    4. Mlika-Cabanne N
    . Clinical practice guidelines for the surgical management of rotator cuff tears in adults. Orthop Traumatol Surg Res. 2010;96:175–179.
    OpenUrlCrossRefPubMed
  15. ↵
    1. Diercks R,
    2. Bron C,
    3. Dorrestijn O,
    4. et al
    ; Dutch Orthopaedic Association. Guideline for diagnosis and treatment of subacromial pain syndrome: a multidisciplinary review by the Dutch Orthopaedic Association. Acta Orthop. 2014;85:314–322.
    OpenUrlCrossRefPubMedWeb of Science
  16. ↵
    1. Ketola S,
    2. Lehtinen J,
    3. Rousi T,
    4. et al
    . No evidence of long-term benefits of arthroscopic acromioplasty in the treatment of shoulder impingement syndrome: five-year results of a randomised controlled trial. Bone Joint Res. 2013;2:132–139.
    OpenUrlAbstract/FREE Full Text
  17. ↵
    1. Kukkonen J,
    2. Joukainen A,
    3. Lehtinen J,
    4. et al
    . Treatment of non-traumatic rotator cuff tears: a randomised controlled trial with one-year clinical results. Bone Joint J. 2014;96-B:75–81.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Moosmayer S,
    2. Lund G,
    3. Seljom US,
    4. et al
    . Tendon repair compared with physiotherapy in the treatment of rotator cuff tears: a randomized controlled study in 103 cases with a five-year follow-up. J Bone Joint Surg Am. 2014;96:1504–1514.
    OpenUrlAbstract/FREE Full Text
  19. ↵
    1. Saltychev M,
    2. Aärimaa V,
    3. Virolainen P,
    4. Laimi K
    . Conservative treatment or surgery for shoulder impingement: systematic review and meta-analysis. Disabil Rehabil. 2014:1–8.
  20. ↵
    1. Colvin AC,
    2. Egorova N,
    3. Harrison AK,
    4. et al
    . National trends in rotator cuff repair. J Bone Joint Surg Am. 2012;94:227–233.
    OpenUrlAbstract/FREE Full Text
  21. ↵
    1. Svendsen SW,
    2. Frost P,
    3. Jensen LD
    . Time trends in surgery for non-traumatic shoulder disorders and postoperative risk of permanent work disability: a nationwide cohort study. Scand J Rheumatol. 2012;41:59–65.
    OpenUrlCrossRefPubMed
  22. ↵
    1. Butler M,
    2. Forte M,
    3. Braman J,
    4. et al
    . Nonoperative and Operative Treatments for Rotator Cuff Tears: Future Research Needs: Identification of Future Research Needs From Comparative Effectiveness Review No. 22 [Internet]. AHRQ Publication No. 13-EHC050-EF. Rockville, MD: Agency for Healthcare Research and Quality. Available at: http://www.ncbi.nlm.nih.gov/books/NBK153178. Accessed June 21, 2015.
  23. ↵
    1. Croft P,
    2. Altman DG,
    3. Deeks JJ,
    4. et al
    . The science of clinical practice: disease diagnosis or patient prognosis? Evidence about “what is likely to happen” should shape clinical practice. BMC Med. 2015;13:20.
    OpenUrlCrossRefPubMed
  24. ↵
    1. Moons KG,
    2. Royston P,
    3. Vergouwe Y,
    4. et al
    . Prognosis and prognostic research: what, why, and how? BMJ. 2009;338:b375.
    OpenUrlFREE Full Text
  25. ↵
    PROGRESS: MRC PROGnosis RESearch Strategy Partnership. Available at: http://www.progress-partnership.org/. 2015. Accessed June 21, 2015.
  26. ↵
    Cochrane Prognosis Methods Group. Available at: http://prognosismethods.cochrane.org/. 2015. Accessed June 21, 2015.
  27. ↵
    1. Hemingway H,
    2. Croft P,
    3. Perel P,
    4. et al
    ; PROGRESS Group. Prognosis Research Strategy (PROGRESS) 1: a framework for researching clinical outcomes. BMJ. 2013;346:e5595.
    OpenUrlFREE Full Text
  28. ↵
    1. Steyerberg EW,
    2. Moons KG,
    3. van der Windt DA,
    4. et al
    . Prognosis Research Strategy (PROGRESS) 3: prognostic model research. PLoS Med. 2013;10:e1001381.
    OpenUrlCrossRefPubMed
  29. ↵
    1. Galea MH,
    2. Blamey RW,
    3. Elston CE,
    4. Ellis IO
    . The Nottingham Prognostic Index in primary breast cancer. Breast Cancer Res Treat. 1992;22:207–219.
    OpenUrlCrossRefPubMedWeb of Science
  30. ↵
    1. Harrell FE
    . Regression Modeling Strategies. New York, NY: Springer; 2001.
  31. ↵
    PROSPERO: International Prospective Register of Systematic Reviews. Available at: http://www.crd.york.ac.uk/PROSPERO/. 2015. Accessed June 21, 2015.
  32. ↵
    1. Braun C,
    2. Bularczyk M,
    3. Heintsch J,
    4. et al
    . Manual therapy and exercises for shoulder impingement revisited. Phys Ther Rev. 2013;18:263–284.
    OpenUrlCrossRef
  33. ↵
    1. Braun C,
    2. Hanchard NC
    . Manual therapy and exercise for impingement-related shoulder pain. Phys Ther Rev. 2010;15:62–83.
    OpenUrlCrossRef
  34. ↵
    1. Geersing GJ,
    2. Bouwmeester W,
    3. Zuithoff P,
    4. et al
    . Search filters for finding prognostic and diagnostic prediction studies in medline to enhance systematic reviews. PLoS One. 2012;7:e32844.
    OpenUrlCrossRefPubMed
  35. ↵
    PROBAST: Prediction Risk of Bias Tool 2015. York, United Kingdom: Kleijnen Systematic Reviews Ltd. 2015. Available at: http://www.systematic-reviews.com/probast. Accessed June 21, 2015.
  36. ↵
    QUADAS-2 2015. University of Bristol, United Kingdom. Available at: http://www.bris.ac.uk/quadas/. 2015. Accessed June 21, 2015.
  37. ↵
    1. Chester R,
    2. Shepstone L,
    3. Daniell H,
    4. et al
    . Predicting response to physiotherapy treatment for musculoskeletal shoulder pain: a systematic review. BMC Musculoskelet Disord. 2013;14:203.
    OpenUrlCrossRefPubMed
  38. ↵
    1. Kuijpers T,
    2. van der Windt DA,
    3. van der Heijden GJ,
    4. Bouter LM
    . Systematic review of prognostic cohort studies on shoulder disorders. Pain. 2004;109:420–431.
    OpenUrlCrossRefPubMedWeb of Science
  39. ↵
    1. Kromer TO,
    2. Sieben JM,
    3. de Bie RA,
    4. Bastiaenen CH
    . Influence of fear-avoidance beliefs on disability in patients with subacromial shoulder pain in primary care: a secondary analysis. Phys Ther. 2014;94:1175–1184.
    OpenUrl
  40. ↵
    1. Hallgren HC,
    2. Holmgren T,
    3. Oberg B,
    4. et al
    . A specific exercise strategy reduced the need for surgery in subacromial pain patients. Br J Sports Med. 2014;48:1431–1436.
    OpenUrlAbstract/FREE Full Text
  41. ↵
    1. Hung CJ,
    2. Jan MH,
    3. Lin YF,
    4. et al
    . Scapular kinematics and impairment features for classifying patients with subacromial impingement syndrome. Man Ther. 2010;15:547–551.
    OpenUrlCrossRefPubMed
  42. ↵
    1. Merolla G,
    2. Paladini P,
    3. Saporito M,
    4. Porcellini G
    . Conservative management of rotator cuff tears: literature review and proposal for a prognostic. Prediction Score. Muscles Ligaments Tendons J. 2011;1:12–19.
    OpenUrlPubMed
  43. ↵
    1. Taheriazam A,
    2. Sadatsafavi M,
    3. Moayyeri A
    . Outcome predictors in nonoperative management of newly diagnosed subacromial impingement syndrome: a longitudinal study. Med Gen Med. 2005;7:63.
    OpenUrl
  44. ↵
    1. Kromer TO,
    2. de Bie RA,
    3. Bastiaenen CH
    . Effectiveness of individualized physiotherapy on pain and functioning compared to a standard exercise protocol in patients presenting with clinical signs of subacromial impingement syndrome: a randomized controlled trial. BMC Musculoskelet Disord. 2010;11:114.
    OpenUrlCrossRefPubMed
  45. ↵
    1. Struyf F,
    2. De Hertogh W,
    3. Gulinck J,
    4. Nijs J
    . Evidence-based treatment methods for the management of shoulder impingement syndrome among Dutch-speaking physiotherapists: an online, Web-based survey. J Manipulative Physiol Ther. 2012;35:720–726.
    OpenUrlCrossRefPubMed
  46. ↵
    1. Varela E,
    2. Valero R,
    3. Küçükdeveci AA,
    4. et al
    ; UEMS-PRM Section Professional Practice Committee. Shoulder pain management: the role of physical and rehabilitation medicine physicians—the European perspective based on the best evidence. A paper by the UEMS-PRM Section Professional Practice Committee. Eur J Phys Rehabil Med. 2013;49:743–751.
    OpenUrlPubMed
  47. ↵
    1. Steyerberg EW,
    2. Harrell FE Jr
    . Prediction models need appropriate internal, internal-external, and external validation. J Clin Epidemiol. 2016;69:245–247.
    OpenUrlCrossRefPubMed
  48. ↵
    1. Siontis GC,
    2. Tzoulaki I,
    3. Castaldi PJ,
    4. Ioannidis JP
    . External validation of new risk prediction models is infrequent and reveals worse prognostic discrimination. J Clin Epidemiol. 2015;68:25–34.
    OpenUrlCrossRefPubMed
  49. ↵
    1. Harrell FE Jr,
    2. Lee KL,
    3. Mark DB
    . Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361–387.
    OpenUrlCrossRefPubMedWeb of Science
  50. ↵
    1. Flom PL,
    2. Cassell DL
    . Stopping stepwise: why stepwise and similar selection methods are bad, and what you should use. Northeast SAS Users Group. Available at: http://www.nesug.org/proceedings/nesug07/sa/sa07.pdf. 2007. Accessed March 22, 2015.
  51. ↵
    1. Deeks JJ,
    2. Bossuyt PM,
    3. Gatsonis C
    1. Reitsma JB,
    2. Rutjes AWS,
    3. Whiting P,
    4. et al
    . Chapter 9: Assessing methodological quality. In: Deeks JJ, Bossuyt PM, Gatsonis C, eds. Cochrane Handbook of Systematic Reviews of Diagnostic Test Accuracy, version 1.0.0. Available at: http://srdta.cochrane.org/.2009. Accessed June 21, 2015.
  52. ↵
    1. Tu YK,
    2. Gilthorpe MS
    . Revisiting the relation between change and initial value: a review and evaluation. Stat Med. 2007;26:443–457.
    OpenUrlCrossRefPubMedWeb of Science
  53. ↵
    1. Vodicka E,
    2. Kim K,
    3. Devine EB,
    4. et al
    . Inclusion of patient-reported outcome measures in registered clinical trials: evidence from ClinicalTrials.gov (2007–2013). Contemp Clin Trials. 2015;43:1–9.
    OpenUrlCrossRefPubMed
  54. ↵
    1. Chatterley T,
    2. Dennett L
    . Utilisation of search filters in systematic reviews of prognosis questions. Health Info Libr J. 2012;29:309–322.
    OpenUrlCrossRefPubMed
  55. ↵
    1. Altman DG
    . Systematic reviews of evaluations of prognostic variables. BMJ. 2001;323:224–228.
    OpenUrlFREE Full Text
  56. ↵
    1. Walker-Dilks C,
    2. Wilczynski NL,
    3. Haynes RB
    . Cumulative Index to Nursing and Allied Health Literature search strategies for identifying methodologically sound causation and prognosis studies. Appl Nurs Res. 2008;21:98–103.
    OpenUrlCrossRefPubMed
  57. ↵
    1. Collins GS,
    2. Reitsma JB,
    3. Altman DG,
    4. et al
    . Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350:g7594–g7594.
    OpenUrlCrossRefPubMed
  58. ↵
    1. Moons KG,
    2. de Groot JA,
    3. Bouwmeester W,
    4. et al
    . Critical appraisal and data extraction for systematic reviews of prediction modelling studies: the CHARMS Checklist. PLoS Med. 2014;11:e1001744.
    OpenUrlCrossRefPubMed
  59. ↵
    1. Liberati A,
    2. Altman DG,
    3. Tetzlaff J,
    4. et al
    . The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions explanation and elaboration. BMJ. 2009:b2700.
View Abstract
Back to top
Vol 96 Issue 7 Table of Contents
Physical Therapy: 96 (7)

Issue highlights

Email
Print
Prognostic Models in Adults Undergoing Physical Therapy for Rotator Cuff Disorders: Systematic Review
Cordula Braun, Nigel C. Hanchard, Alan M. Batterham, Helen H. Handoll, Andreas Betthäuser
Physical Therapy Jul 2016, 96 (7) 961-971; DOI: 10.2522/ptj.20150475

Citation Manager Formats

Download Powerpoint
Save to my folders

Prognostic Models in Adults Undergoing Physical Therapy for Rotator Cuff Disorders: Systematic Review
Cordula Braun, Nigel C. Hanchard, Alan M. Batterham, Helen H. Handoll, Andreas Betthäuser
Physical Therapy Jul 2016, 96 (7) 961-971; DOI: 10.2522/ptj.20150475
del.icio.us logo Digg logo Reddit logo Technorati logo Twitter logo CiteULike logo Connotea logo Facebook logo Google logo Mendeley logo

Subjects