Read on Twitter
1/
Our new Individual participant data (IPD) meta-analysis of prognostic factors for anterior knee pain in adolescents published in @PAINthejournal. Thread: process, results & reflections
#IPD #prognosis #OPTIYOUTH
LINK: https://bit.ly/3a1prfv @MichaelRathleff
Our new Individual participant data (IPD) meta-analysis of prognostic factors for anterior knee pain in adolescents published in @PAINthejournal. Thread: process, results & reflections
#IPD #prognosis #OPTIYOUTH
LINK: https://bit.ly/3a1prfv @MichaelRathleff
What is an IPD? IPD meta-analysis synthesises raw data from multiple studies. This has been championed as the gold-standard for synthesising prognostic factor studies; read more here: https://pubmed.ncbi.nlm.nih.gov/22530717/ ; @Richard_D_Riley and: https://bjsm.bmj.com/content/51/20/1456 @BJSM_BMJ
3/ WHY IPD? (some reasons)
1.Include prospective studies which never evaluated prognostic factors
2.Include analyses/outcomes not included in original studies-> >sample size
3.Apply inclusion at individual level
@NatJCollins @JillAHayden
1.Include prospective studies which never evaluated prognostic factors
2.Include analyses/outcomes not included in original studies-> >sample size
3.Apply inclusion at individual level
@NatJCollins @JillAHayden
4/ WHY IPD (CTD)
4.Harmonise data: > pooling which is difficult in prognostic studies due to variability in definitions/scales of variables (e.g. https://pubmed.ncbi.nlm.nih.gov/27525834/ )
5.Standardise analyses
+more reasons!
@NatJCollins @JillAHayden @Richard_D_Riley
4.Harmonise data: > pooling which is difficult in prognostic studies due to variability in definitions/scales of variables (e.g. https://pubmed.ncbi.nlm.nih.gov/27525834/ )
5.Standardise analyses
+more reasons!
@NatJCollins @JillAHayden @Richard_D_Riley
5/ So HOW did we do it? Our IPD was prospectively registered( https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42019116861). Primary outcomes: long-term (12months) pain (intensity) and disability (KOOS Sport/rec). Below demonstrate process for identifying studies.
6/ We then had to code and harmonise datasets for analyses. This means standardising scales and measurement of variables to maximise data availability. Data checks & validations were done before, during & after.
7/ We included a broad sample of participants and were able to describe them at the individual level on important characteristics
8/ We showed that the propensity for chronicity was high-for both pain and disability. Most improvements occurred in the short-term, with limited changes in the longer term. So maybe not as self-limiting as once assumed?
9/ Knee strength, BMI or pain sensitivity were NOT associated with long term pain OR disability
10/ Pain severity (> frequency, intensity, duration) & < quality of life were associated with > long term pain and disability. Moderate anxiety/depression was associated with greater disability after 12months
9/ CAVEAT: some factors were not clinically relevant e.g. pain duration, +1year was associated with 2pt (/100) >pain & disability – so only with >5-10 years pain duration would this be clinically important. Is risk stratification realistic? My bias is no…
10/ Despite including and pooling the largest sample of adolescents to date, we are a long way from robust findings on strong prognostic factors..Recent lit on back pain shows stratified treatment may not be effective..
11/ But we do plan to find out in this population ! Work in progress lead by @Andreucci_Ale to develop the AK- Pain prognostic tool to see if these (&other) factors can identify clinically meaningful sub-groups. (see: https://f1000research.com/articles/8-2148/v1)
11/ Acknowledgements – IPD is notoriously more time and labour intensive that standard IPD and cannot be don’t without a big network of collaborators willing to share data. Very grateful to work with the great adolescent knee health group members on this
