Duke Doctor of Physical Therapy - The Impact of ... › files › Team_5_Manual_Therapy...bias...

1
Option 2 studies consistently produced larger SMDs than Option 1; Option 3 not present in each stratification for analysis. PEDro scores for Option 3 (7.2 + 1.1) significantly higher than Option 1 (6.4 + 1.2, p=0.03) and Option 2 (6.2 + 1.4, p = .003) Single- mode interventions (e.g. mobilization/manipulation) more likely to be performed in Option 1 studies (p = .01) Publication trends show an increase of Option 1 and 2 studies and a decrease in Option 3 studies Background & Purpose The Impact of Measurement Bias on Effect Size in Manual Therapies of the Spine Andrea Rawley, SPT, Meghan O’Hara, SPT, Kimberly Kurtz, SPT, Leigh Martino, SPT, Lisa Cole, SPT, Matthew O’Connell, SPT, Joshua Staggs, SPT, Jack Friesen, SPT, Michael P. Reiman, PT, DPT, OCS, SCS, ATC, FAAOMPT, CSCS, Charles Sheets, PT, OCS, SCS, Dip MDT Mobilizations and manipulations are commonly used treatments in the management of back pain. The effectiveness of these skilled interventions is dependent on the expertise of the treating clinician, creating potential for bias. Quantifying differences in pain outcomes and internal validity based on the number of treating clinicians and the number of interventions they each perform will give better understanding to this potential bias. This systematic review aims to examine bias and internal validity within randomized controlled trials (RCTs) utilizing mobilization and manipulation of the spine. Hypothesis: Single clinician studies will show inflated effect sizes for pain outcomes and greater risk to internal validity. RCTs from PubMed and Cochrane databases identified and included or excluded based on review criteria. Each study categorized by option choice. Data extraction performed and cross checked by all authors. PEDro scores pulled when available, or scored by CS. Standard mean differences (SMDs), 95% confidence intervals (CIs) calculated from extracted data with R Project for Statistical Computing (stratified by region of spine and follow up time frame) Analysis of PEDro scores performed using one-way ANOVA Chi squared tests to compare experimental interventions within option choices Examination of publication trends over time Single therapist bias does not appear to drive pain outcomes Differences between option choices are present and significant regarding internal validity, type of intervention, and publication rate; differences not in favor of any particular option choice Extreme heterogeneity makes any extrapolation challenging Additional research needed to better understand this potential bias within physical therapy studies PEDro Scores Stratified by Option Choice Pain Score SMDs Stratified by Follow Up Time & Region of Spine Option n SMD (95% CI) Immediate Cervical 1 9 0.32 (-.22, .87) 2 4 0.67 (.25, 1.08) 3 2 0.16 (-.12, .45) Non-immediate Cervical 1 11 0.11 (-.27, .50) 2 19 0.34 (.05, .62) 3 13 0.59 (.21, .97) Immediate Lumbar 1 4 0.32 (.07, .56) 2 4 0.60 (-.07, 1.28) Non-immediate Lumbar 1 6 0.20 (-.09, .48) 2 28 0.34 (.10, .58) 3 15 0.23 (.03, .43) Immediate Cervical Non-immediate Lumbar Non-immediate Cervical Immediate Lumbar Option Choices Option 1 One clinician performing experimental and control treatment interventions Option 2 Multiple clinicians performing one treatment intervention only Option 3 Multiple clinicians performing experimental and control treatment interventions We thank our mentors Charles Sheets and Dr. Michael Reiman for their guidance and expertise in clinical research. We also thank Leila Ledbetter for her assistance with our searches and her expertise in search engines and open-source reference management software. Methods Analysis Results Conclusions & Clinical Relevance Acknowledgements

Transcript of Duke Doctor of Physical Therapy - The Impact of ... › files › Team_5_Manual_Therapy...bias...

Page 1: Duke Doctor of Physical Therapy - The Impact of ... › files › Team_5_Manual_Therapy...bias within physical therapy studies PEDroScores Stratified by Option Choice Pain Score SMDs

• Option2studiesconsistentlyproducedlargerSMDsthanOption1;Option3notpresentineachstratificationforanalysis.

• PEDroscoresforOption3(7.2+ 1.1)significantlyhigherthanOption1(6.4+ 1.2, p=0.03)andOption2(6.2+ 1.4,p=.003)

• Single- modeinterventions(e.g.mobilization/manipulation)morelikelytobeperformedinOption1studies(p=.01)

• PublicationtrendsshowanincreaseofOption1and2studiesandadecreaseinOption3studies

Background&Purpose

TheImpactofMeasurementBiasonEffectSizeinManualTherapiesoftheSpine

AndreaRawley,SPT,MeghanO’Hara,SPT,KimberlyKurtz,SPT,LeighMartino,SPT,LisaCole,SPT,MatthewO’Connell,SPT,JoshuaStaggs,SPT,JackFriesen,SPT,MichaelP.Reiman,PT,DPT,OCS,SCS,ATC,FAAOMPT,CSCS,CharlesSheets,PT,OCS,SCS,DipMDT

Mobilizationsandmanipulationsarecommonlyusedtreatmentsinthemanagementofbackpain.Theeffectivenessoftheseskilledinterventionsisdependentontheexpertiseofthetreatingclinician,creatingpotentialforbias.Quantifyingdifferencesinpainoutcomesandinternalvaliditybasedonthenumberoftreatingcliniciansandthenumberofinterventionstheyeachperform willgivebetterunderstandingtothispotentialbias.Thissystematicreviewaimstoexaminebiasandinternalvaliditywithinrandomizedcontrolledtrials(RCTs)utilizingmobilizationandmanipulationofthespine.Hypothesis:Singleclinicianstudieswillshowinflatedeffectsizesforpainoutcomesandgreaterrisktointernalvalidity.

RCTsfromPubMedandCochranedatabasesidentifiedandincludedorexcludedbasedonreviewcriteria.Eachstudycategorizedbyoptionchoice.Dataextractionperformedandcrosscheckedbyallauthors.PEDroscorespulledwhenavailable,orscoredbyCS.

• Standardmeandifferences(SMDs),95%confidenceintervals(CIs)calculatedfromextracteddatawithRProjectforStatisticalComputing(stratifiedbyregionofspineandfollowuptimeframe)

• AnalysisofPEDroscoresperformedusingone-wayANOVA• Chisquaredteststocompareexperimentalinterventionswithin

optionchoices• Examinationofpublicationtrendsovertime

• Singletherapistbiasdoesnotappeartodrivepainoutcomes• Differencesbetweenoptionchoicesarepresentandsignificant

regardinginternalvalidity,typeofintervention,andpublicationrate;differencesnotinfavorofanyparticularoptionchoice

• Extremeheterogeneitymakesanyextrapolationchallenging• Additionalresearchneededtobetterunderstandthispotential

biaswithinphysicaltherapystudies

PEDro ScoresStratifiedbyOptionChoice

Pain ScoreSMDsStratifiedbyFollowUpTime&RegionofSpine

Option n SMD(95%CI)ImmediateCervical

1 9 0.32 (-.22,.87)2 4 0.67(.25,1.08)3 2 0.16 (-.12,.45)

Non-immediateCervical1 11 0.11 (-.27,.50)2 19 0.34 (.05,.62)3 13 0.59 (.21,.97)

ImmediateLumbar1 4 0.32 (.07,.56)2 4 0.60(-.07,1.28)

Non-immediateLumbar1 6 0.20(-.09,.48)2 28 0.34(.10,.58)3 15 0.23(.03,.43)

ImmediateCervical

Non-immediateLumbarNon-immediateCervical

ImmediateLumbar

OptionChoicesOption1 One clinicianperformingexperimentalandcontroltreatmentinterventions

Option2 Multiple cliniciansperformingonetreatmentinterventiononly

Option3 Multiple cliniciansperformingexperimentalandcontroltreatmentinterventions

WethankourmentorsCharlesSheetsandDr.MichaelReimanfortheirguidanceandexpertiseinclinicalresearch.WealsothankLeilaLedbetterforherassistancewithoursearchesandherexpertiseinsearchenginesandopen-sourcereferencemanagementsoftware.

Methods

Analysis

Results

Conclusions&ClinicalRelevance

Acknowledgements