Journals LibraryNHS NIHR - National Institute for Health Research
Rethinking specific disease labels: Response to primary care interventions across different pain syndromes – an individual patient data meta-analysis
Rethinking specific disease labels: Response to primary care interventions across different pain syndromes - an individual patient data meta-analysis
27 April 2015
01 January 2011
31 July 2014
3 years 7 months
Pain; Prognosis; Randomised Clinical Trials; Primary Care
- Professor Danielle van der Windt, Arthritis Research UK Primary Care Centre, Keele University
- Dr Majid Artus, Arthritis Research UK Primary Care Centre, Keele University
- Dr Martyn Lewis, Arthritis Research UK Primary Care Centre, Keele University
- Professor Rona Moss-Morris, School of Psychology, University of Southampton
- Dr Marta Buszewicz, Research Department of Primary Care & Population Health, UCL
This is a small project based on secondary analysis of existing trial data. The project as such has not led to new collaborations forged outside of the NIHR SPCR. However, the research into prognosis and stratified care more generally has led to many new research collaborations. For example, Keele researchers contribute to the MRC PROGRESS Partnership for prognosis research, led by Professor Harry Hemingway – UCL, with additional partners from the University of Birmingham (Professor Richard Riley, appointed at Keele University in 2014), the London School of Hygiene and Tropical Medicine (Professor Ian Roberts, Dr Pablo Perel), and the University of Oxford (Professor Doug Altman). The stratified care team collaborates with several groups who are developing trials to investigate the effectiveness of stratified care in musculoskeletal pain, e.g. in the US (Dr Dan Cherkin, University of Washington) and in the Netherlands (Professor Raymond Ostelo, VU University Amsterdam).
The overall aim of this project is to test the observation that patients with different types of pain syndromes show a similar pattern of response to primary care treatments, and that similar factors predict a poor outcome of treatment. Emphasis will be put on factors that may be modified by treatment, such as baseline intensity of pain, psychological distress, presence of multiple pains, or inadequate pain attributions and beliefs. Key objectives are:
- to investigate similarities and differences in baseline characteristics of trial participants with either neck pain, shoulder pain, low back pain, tennis elbow or knee pain;
- to investigate similarities and differences in the response to treatment (recovery rate, improvement of pain and disability) at short-term (6-12 weeks) and long-term (6-12 months) follow-up across trial participants with different types of pain;
- to explore differences and similarities in the predictive value of potential prognostic indicators across trial participants with different types of pain syndromes.
Changes to project objectives
There have been no changes to the original proposal.
The majority of studies on the clinical course, prognosis and management of musculoskeletal pain problems have focussed on a specific or regional pain condition, such as back pain, shoulder pain or tennis elbow. The results of many of these studies show similar findings in terms of clinical characteristics, symptom trajectories, and prognostic factors. Previous research has also highlighted factors that consistently predict outcome regardless of pain site, including socioeconomic variables, severity and history of pain, physical functioning, and psychological factors. Mallen et al.  recently demonstrated that a brief set of generic prognostic indicators (duration of present pain episode, pain interference with daily activities, and presence of multiple-site pain) improved on clinicians’ estimates of prognosis in older patients with a range of musculoskeletal conditions presenting in primary care.
These findings support the hypothesis that in patients with musculoskeletal pain, generic factors are more important in the prediction of future outcome (prognosis) than the location or patho-anatomic cause of pain. This may also generate opportunities for the design and evaluation of interventions that target potentially modifiable prognostic factors regardless of pain location. The objective of this study was to investigating the course of pain and disability in trial participants with different regional pain problems, identify generic predictors of outcome (prognostic factors), and explore to what extent the association between predictors and outcome in trial participants is modified by pain location.
This study was based on secondary analysis of individual patient data from seven randomised clinical trials carried out within the Arthritis Research UK Primary Care Centre (Keele University, UK), investigating a range of primary care interventions, and published between 2004 and 2011. The outcome measures for all analyses were (1) pain intensity and (2) functional disability. In all datasets the scores for pain and disability were transformed onto a 0-10 scale with higher scores indicating more severe pain or disability, to allow comparison of descriptive results across studies and pooling of data for analysis. Potential predictors were identified from each dataset, and recoded where needed to ensure consistency between datasets. Variables included for analysis were age, sex, social class (Manual, Non-Manual), presence of widespread pain according to ACR-90 criteria, duration of pain episode, baseline pain score, baseline disability score, and mood problems (item from EQ-5D).
Analysis of the effectiveness of specific interventions was not an objective of this study, and therefore interventions were classified into either active, control or placebo depending on the nature of the treatment. Changes from baseline at short and long-term follow-up were calculated for both pain and disability. Linear regression was used to analyse changes in pain and functional disability across different pain regional pain conditions. Baseline values of pain and disability were entered as these are often found to be the strongest predictor of outcome. Additional potential predictors were then included to investigate which variables generically predicted outcome regardless of pain location. Type of treatment (active, placebo or control) was included as a potential confounder. The percentage explained variance was calculated to quantify the importance of each predictor by multiplying the beta coefficient of each predictor by the correlation of the predictor with outcome (pain or disability). Interaction terms (predictor*pain location) were finally included to explore to what extent the association between prognostic factors and treatment outcome varied between different pain locations.
The number of participants from each respective trial ranged from 164 to 851, resulting in a total sample of 2,651 participants, with 2,483 providing data for analysis.
Course of symptoms
The results show considerable variation between trials in the level of pain and disability, yet in all trials most improvement of symptoms occurred over the first three months, with little further change over subsequent 3-18 months follow-up. Participants of the back pain trials showed larger short-term improvements compared to all other pain regions. Using back pain as the reference, the mean pain intensity score at 12 months (adjusted for differences in baseline values) was almost 2 points higher for participants with knee pain compared to back pain (mean difference: -1.85, 95% CI -2.12, -1.57).
Increasing age, longer pain duration, manual work, presence of widespread pain, and mood problems (moderate or extreme anxiety or depressive symptoms) were consistently associated with poor outcome, also when adjusted for pain location or the type of treatment received. As expected, baseline outcome scores explained most of the variance in the same outcome at follow-up (>58%). Otherwise, pain duration, age, and location (in particular knee pain) contributed most to the prediction of changes in pain (explained variance 4.13 to 13.98%) or disability (explained variance 6.13% to 16.75%).
Variation in prognostic factors across pain sites
Significant interactions nearly all concerned location of pain at the knee. Participants with knee pain showed stronger associations with poor outcome for baseline levels of pain and disability, increasing age, manual work, and longer pain duration. Few additional interactions were found. No significant interactions were found for widespread pain and mood problems, indicating that these variables had a similar effect on outcome across all pain locations.
Differences in prognostic factors and long-term outcome were mostly found for participants of knee pain trials who were older and had more chronic conditions (osteoarthritis). Similarity of predictors of treatment outcome across other regional pain problems provides evidence to support targeting of treatment based on prognostic factors rather than location of pain.
Plain English summary
Results of existing research shows that different types of pain, such as back pain, tennis elbow, or shoulder pain, often have similar features. For example, the course of pain over time and the impact of pain on everyday life is similar, regardless of where the pain is. Furthermore, factors such as the duration and severity of the pain, and whether or not patients have mood problems, has a similar influences the course of the problem, again regardless of where the pain is. The aim of this study was to investigate the course of pain and disability following treatment in patients with different pain problems participating in clinical trials, and investigate the similarities and differences in factors predicting future outcome.
Data were used from seven clinical trials investigating a range of primary care treatments for tennis elbow, neck, shoulder, back, or knee pain. Measures of outcomes were severity of pain and disability. The course of pain and disability over 12 months follow-up was described. We then analysed whether the type or location of pain influenced changes in pain and disability over time. We also compared how well factors like age, gender, social class, pain duration, widespread pain, and mood problems, predicted future changes in pain and disability, and if this different between types of pain.
Data from a total of 2,483 patients were included in the study. The results showed a similar course of pain and disability over time, regardless of pain location, with a quick improvement over the first 3 months of the trial, after which little further change occurred. The only exception was seen for participants with knee pain, who on average showed a poorer long-term outcome. More widespread pain and increasing levels of anxiety or depression predicted a poorer outcome in participants with any type of pain. Especially in participants with knee pain, increasing age, manual work and long pain duration were strong predictors of poor outcome.
We found that predictors of the outcome of treatment are often the same in patients with musculoskeletal pain problems, no matter where the pain is. This means that it is important to take these factors (such as the presence of widespread pain or mood problems) into account when treating different types of musculoskeletal pain problems.
- Hayden JA, van der Windt DA, Cartwright JL, Côté P, Bombardier C. Assessing bias in studies of prognostic factors. Ann Intern Med. 2013 Feb 19;158(4):280-6.
- Steyerberg EW, Moons KG, van der Windt DA, Hayden JA, Perel P, Schroter S, Riley RD, Hemingway H, Altman DG; PROGRESS Group. Prognosis Research Strategy (PROGRESS) 3: prognostic model research. PLoS Med. 2013 Feb;10(2):e1001381.
- Riley RD, Hayden JA, Steyerberg EW, Moons KG, Abrams K, Kyzas PA, Malats N,Briggs A, Schroter S, Altman DG, Hemingway H; PROGRESS Group. Prognosis Research Strategy (PROGRESS) 2: prognostic factor research. PLoS Med. 2013 Feb;10(2):e1001380.
- Hemingway H, Croft P, Perel P, Hayden JA, Abrams K, Timmis A, Briggs A, Udumyan R, Moons KG, Steyerberg EW, Roberts I, Schroter S, Altman DG, Riley RD; PROGRESS Group. Prognosis research strategy (PROGRESS) 1: a framework for researching clinical outcomes. BMJ. 2013 Feb 5;346:e5595.
- Hingorani AD, van der Windt DA, Riley RD, Abrams K, Moons KG, Steyerberg EW, Schroter S, Sauerbrei W, Altman DG, Hemingway H; PROGRESS Group. Prognosis research strategy (PROGRESS) 4: stratified medicine research. BMJ. 2013 Feb 5;346:e5793.
- Croft P, Altman DG, Deeks JJ, Dunn KM, Hay AD, Hemingway H, LeResche L, Peat G, Perel P, Petersen SE, Riley RD, Roberts I, Sharpe M, Stevens RJ, Van Der Windt DA, Von Korff M, Timmis A. The science of clinical practice: disease diagnosis or patient prognosis? Evidence about "what is likely to happen" should shape clinical practice. BMC Med. 2015 Jan 30;13(1):20.
This project is based on secondary analysis of existing data and has been conducted without PPI. The grants and projects developed from this work (see impact section below) do have strong PPI.
The results of this work have mainly been used in the development of further grant applications and projects regarding the prognosis of musculoskeletal problems. For example, the results of this exploratory study have been valuable in the design of a programme of work (including our funded NIHR STarT-Musc Programme for Applied Research) to investigate predictors of outcome across musculoskeletal conditions, and develop a model of stratified care for musculoskeletal pain based on prognostic stratification with matched treatment options.
The results of this project have also contributed to the design of several prognosis studies funded by SPCR (e.g. 209/210: design and testing of a Smartphone & Tablet application to record short-term pain trajectories in patients with painful musculoskeletal conditions; and 187: using longitudinal data in prognosis research). All these studies investigate the predictive value and clinical usefulness of prognostic factors in patients with musculoskeletal conditions, regardless of the type or location of pain.
The work has been prepared for publication as a journal article (not yet available in the public domain).
This project was funded by the National Institute for Health Research School for Primary Care Research (project number 84)
Department of Health Disclaimer
The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR School for Primary Care Research, NIHR, NHS or the Department of Health.