Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Bias in cluster trials
- David Torgerson (21 June 2005)
-
Re: Bias in cluster trials
- Danielle Van der Windt, Petra Jellema (29 June 2005)
-
Inadequate training?
- John F Morgan (8 July 2005)
-
Psychosocial concerns - a matter for question asking?
- Wolf Langewitz (11 July 2005)
-
Don't over simplfy
- Richard Colman (12 July 2005)
-
Psychological risk factors for chronicity are not well known
- Tamar Pincus, Alan Breen, Kim Burton, Martin Underwood (13 July 2005)
-
Re: rapid responses
- Petra Jellema, Danielle van der Windt (22 July 2005)
-
Intervention and misclassification
- Peter Vedsted, Kaj S. Christensen (25 July 2005)
|
|
|||
|
David Torgerson, Director, York Trials Unit Dept of Health Sciences, University of York YO10 5DD
Send response to journal:
|
The paper by Jellema and colleagues used cluster randomisation without pre-identification of participants. This approach invites selection bias in cluster trials [1,2]. GPs were trained in the management of back pain and then were asked to identify participants. This approach prevents allocation concealment and the training itself may change GP behaviour in the method of participant identification. A similar approach was used in a pilot study for the MRC BEAM trial for treatments of back pain. The pilot showed significant identification bias, leading to the abandonment of cluster allocation for the main trial [3]. Is there evidence for problems in this trial? The authors do not give the GP practice size (which has been recommended for cluster trials [2]) so it is difficult to estimate the recruitment rates between the two group. But assuming randomisation equaled the practice sizes we can see that the trained practices recruited on average 5.3 participants each compared with the control of 6.2. More worryingly, however, is that after recruitment 14% (n = 27) of the control group were excluded compared with only 3% (n = 5) of the intervention group. These post randomisation exclusions were they done blindly - if not this could be a source of bias. Ideally, the authors should have identified prevalent back pain patients before random allocation and training of the GPs. This would have then ensured no chance of recruitment or exclusion bias. This trial, unfortunately, like many cluster trials published in major medical journals has evidence for exclusion and possibly recruitment bias [2]. David Torgerson, Director York Trials Unit. References [1]Hahn S, Puffer S, Torgerson DJ, Watson J. Methodological bias in cluster randomised trials. BMC Medical Research Methodology 2005;5:10. [2] Puffer S, Torgerson DJ, Watson J. Evidence for risk of bias in cluster randomised trials: review of recent trials published in three general medical journals. BMJ 2003;327:785. [3]Farrin A, Russel I, Torgerson D, Underwood M. Differential recruitment in cluster randomized trial in primary care: the experience of the UK Back pain, Exercis, Active management and Manipulation (UK BEAM) feasibility study. Clinical Trials 2005;2:119-24. Competing interests: None declared |
|||
|
|
|||
|
Danielle Van der Windt, senior researcher VUmc medical centre, EMGO Institute, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands, Petra Jellema
Send response to journal:
|
We would like to thank David Torgerson for his comments on our cluster-randomised trial among patients with acute or subacute low back pain (LBP). Torgerson raises the important issue of selection bias in trials in which randomisation takes place at the level of the care provider, while the care providers themselves are responsible for patient recruitment. After randomisation, practice size was more or less equally distributed between our two intervention groups: mean (SD) number of registered patients were 2619 (370) and 2700 (523). But indeed, the general practitioners (GPs) in the usual care (UC) group recruited more patients than the GPs in the minimal intervention strategy (MIS) group. We believe that this difference was mainly explained by the fact that recruitment of patients in the MIS group entailed a lot more work for GPs than in the UC group. There is little evidence that this difference in recruitment rate has resulted in selection bias. Table 1 of our paper shows that the two patients groups were quite similar at baseline with respect to prognostic indicators and most baseline values of outcome measures. Furthermore, our population was largely comparable to (sub)acute LBP populations described in other studies in general practice [1,2,3]. Torgerson is absolutely right when he suggests that the optimal method for selecting patients in cluster-randomised trials is to select patients first (e.g. based on patient records), and subsequently randomise and train GPs. This was, however, not feasible in this trial as we aimed for a population of patients who consulted their GP for (sub)acute LBP. Prior identification of patients who will consult their GP in future is, of course, not possible. An alternative to prior identification is identification and recruitment of participants by someone blinded to the group allocation [4]. In a study like ours this implies that a blinded, independent observer has to attend all GP consultations to identify LBP patients fitting the inclusion criteria. We did not have the resources for such a large undertaking. Torgerson correctly states that more patients in the UC group were excluded than in the MIS group. Exclusion of patients was based on pre- defined exclusion criteria and were carried out by the research team. Most exclusions concerned current treatment of patients (e.g. physiotherapy, which was not allowed), age (not between 18 and 65), or the fact that symptoms had already resolved. These exclusions are unlikely to result in bias. We do recognise the risks of selection and exclusion bias in cluster- randomised trials. However, given adequate baseline similarity, and the fact that exclusions were based on objective criteria, it seems highly unlikely that our findings (very small and non-significant difference between intervention groups) can be explained by these biases. REFERENCES 1. Schers HJ, Braspenning JC, Drijver R, Wensing M, Grol RP. Low back pain in general practice: reported management and reasons for not adhering to the guidelines in the Netherlands. British Journal of General Practice 2000;50:640-644. 2. Engers AJ, Wensing M, van Tulder MW, Timmermans A, Oostendorp RA, Koes BW, Grol RP. Implementation of the Dutch low back pain guideline for general practitioners: a cluster randomized controlled trial. Spine 2005;30:595-600. 3. Hay EM, Mullis R, Vohora K, Main CJ, Watson P, Dziedzic KS, Sim J, Minns Lowe C, Croft PR. Comparison of physical treatments versus a brief pain-management programme for back pain in primary care: a randomised clinical trial in physiotherapy practice. Lancet 2005;365:2024-2030. 4. Puffer S, Torgerson DJ, Watson J. Evidence for risk of bias in cluster randomised trials: review of recent trials published in three general medical journals. BMJ 2003;327:785-789. Competing interests: None declared |
|||
|
|
|||
|
John F Morgan, Senior Lecturer in Psychological Medicine St George's Hospital, London
Send response to journal:
|
Sir, Jellema et al provide a rather bleak message over the role of psychosocial interventions in lower back pain. In a well-designed study, they rightly defend the short consultation times (as little as 20 minutes) as representative of the pragmatics of primary care. But there are many effective treatments which primary care practioners have neither the time nor the training to provide. That is the role of the specialist clinic in secondary care and, considering the fiscal and societal burden of lower back pain, the lack of ready access to such clinics is shocking. Liaison (general hospital) psychiatrists provide cognitive behaviour therapy and other brief psychological interventions to thousands of back pain suffers each year in the UK. To suggest that non-specialists given 5 hours training fail to produce a significant response from a treatment lasting 20 minutes is hardly surprising. Perhaps we might next learn about the ineffectiveness, say, of microbiologists trained for 5 hours in carrying out coronary artery bypass grafts within a third of the usual cardiologists' time? Psychological interventions are more complex than CABGs, but there is a current trend to assume they can be manualised, administered by non- specialists and 'done on the cheap'. This study simply demonstrates that this fails. Yours faithfully John F Morgan Competing interests: None declared |
|||
|
|
|||
|
Wolf Langewitz, Head, Dept. Psychosomatic Medicine CH 4031 Basel University Hospital
Send response to journal:
|
Dear Editor The paper by Jellema et al. aims at the identification of the psychosocial burden in patients with lower back pain. The identification strategy was a mixture of asking (open) questions and intensifying the search when patients related material which 'gave the impression that this factor could be an obstacle to recovery'. If we assume that the goal of such an intervention is to reach a common reality between patient and physician this may not be a successful strategy. After such a series of questions and answers we may know something about psychosocial risk factors but little about their contribution to the individual patient's view of his or her personal situation. Instead of using a physician-centred approach to construct a list of psychosocial risk factors a patient-centred approach to building a common model of understanding would be more promising. However, this is a difficult communication task. Given the literature on communication skills training (1) two 2.5 hours training sesssion per GP are not enough to enable them to elicit the patients' model in 20 minutes consultation time. Thus, the study seems to be of the type MISSION IMPOSSIBLE where achievements call for other heroes than BMJ contributors and readers. Wolf Langewitz, Head Dept. Psychosomatic Medicine /Internal Medicine 1. Lewin SA, Skea ZC, Entwistle V, Zwarenstein M, Dick J. Interventions for providers to promote a patient-centred approach in clinical consultations. The Cochrane Database of Systematic Reviews 2001, Issue 4. Art. No.: CD003267. DOI: 10.1002/14651858.CD003267. Competing interests: None declared |
|||
|
|
|||
|
Richard Colman, Occupational physician Industrial doctors teesside. TS23 1PZ
Send response to journal:
|
To me this paper tells us that 20 mins intervention by some GPs is ineffective in influencing sub acute low back pain. I am not surprised at this. It does not say that to effectively address psycho-social factors is a waste of time Competing interests: None declared |
|||
|
|
|||
|
Tamar Pincus, Reader in psychology Royal Holloway University of London TW20 0EX, Alan Breen, Kim Burton, Martin Underwood
Send response to journal:
|
Jellema et al failed to show a clinical benefit from adding a minimal intervention aimed at ‘identification and discussion of psychosocial prognostic factors’ to ‘usual care’ for sub-acute low back pain. Their findings do not demonstrate that a ‘psychosocial’ approach is ineffective. They suggest that we need to look again at the range (and combinations) of physical, psychological, social and societal factors that predict a poor prognosis. then we can develop, and test, potentially effective, more intensive, interventions for those people with acute / sub-acute low back pain most likely to develop chronic disability. The only evidence-based source quoted in the intervention’s development is a systematic review of psychological risk factors for poor prognosis in acute low back pain1. This found weak evidence for ‘catastrophising’ and no evidence for ‘fear-avoidance’ as predictors of poor prognosis in patients with acute/sub-acute low back pain. ‘Distress’ was the only factor shown to be associated with a poor prognosis. The main foci of Jellema et al’s intervention were ‘catastrophising’ and ‘fear -avoidance’, ‘distress’ was not addressed. The absence of evidence for the role of ‘catastrophising’, ‘fear avoidance’ and other putative psychological and other risk factors may be due to the low quality and small size of existing inception cohort studies. Furthermore, the effect of interactions between psychological risk factors and others associated with the health care system in which individuals are treated is unknown. The Multinational Musculoskeletal Inception Cohort project (www.mmics.org) is a collaboration between expert teams in twelve nations and their respective care-system representatives. This project is producing an international evidence-based protocol to allow for pooling data from prospective cohorts across health and social systems. Until MMICS is complete, we cannot assume that we know the important risk factors for poor prognosis, how we should identify them or intervene to change them. 1 Pincus T, Burton K, Vogel S & Field A., A systematic review of psychological risk factors for chronicity/disability in prospective cohorts of low back pain, Spine, on -line Issue, 2002, March 27(5), E109- E120. Competing interests: Members of the MMICS UK team |
|||
|
|
|||
|
Petra Jellema, researcher VUmc medical center, EMGO-Institute, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands, Danielle van der Windt
Send response to journal:
|
We are grateful for the responses to our paper on the cluster- randomised trial among patients with acute or subacute low back pain, and like to comment briefly on each of these responses. We would like to thank John F Morgan and Wolf Langewitz for pointing at the important distinction between treatments aimed at psychosocial factors in primary and secondary care. Each health care system has its own professionals, own treatments and own patient population. While a minimal intervention strategy of 20 minutes will be regarded as a ‘mission impossible’ in a secondary health care setting, this is not by definition the case in primary health care, as most patients will present far less complex LBP episodes than in secondary care. Our five hours of training were therefore not meant to make specialists out of non-specialists. This also means that our conclusions only apply to ‘our’ type of intervention (which we never referred to as a psychological or a cognitive-behavioural intervention) in a general practice setting. Tamar Pincus and colleagues state that catastrophising and fear- avoidance were the main foci of our intervention and that distress was not addressed. This is, however, not correct. The topic ‘worries about the pain’ [Box 1] included questions on psychological distress and eventually depressive symptoms. Furthermore, Tamar Pincus and colleagues question the evidence for the predictive value of the psychosocial variables addressed in our study. They describe the evidence for catastrophising as ‘weak’, and for fear- avoidance as ‘no evidence’. However, in their systematic review Pincus et al. [1] concluded that future research should concentrate on the more promising psychological factors, such as fear avoidance and catastrophising [1;2]. This conclusion actually partly contributed to the design of our intervention. Also other reviews have found evidence for the role of pain catastrophising and fear-avoidance as predictors of poor prognosis [3;4;5]. Consequently, we have confidence that the factors we chose to address in our study do matter in the transition from acute to chronic low back pain. We agree with Richard Colman that the fact that we found no effectiveness of our intervention does not say that it is a waste of time to effectively address psychosocial factors. Our study only shows that our minimal intervention strategy is probably not the best way to address these factors effectively. REFERENCES [1] Pincus T, Burton AK, Vogel S, Field AP. A systematic review of psychological factors as predictors of chronicity/disability in prospective cohorts of low back pain. Spine 2002a;27:E109-120. [2] Pincus T, Vlaeyen JWS, Kendall NAS, Von Korff MR, Kalauokalani DA, Reis S. Cognitive-behavioral therapy and psychosocial factors in low back pain; directions for the future. Spine 2002b;27:E133-138. [3] Linton SJ. A review of psychological risk factors in back and neck pain. Spine 2000;25:1148-1156. [4] Vlaeyen JWS, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. Pain 2000;85:317-332. [5] Waddell G, Burton AK, Main CJ. Screening to identify people at risk of long-term incapacity for work; a conceptual and scientific review. London: Royal Society of Medicine Press Ltd, 2003. Competing interests: None declared |
|||
|
|
|||
|
Peter Vedsted, Senior researcher Research Unit for General Practice, University of Aarhus, Vennelyst Boulevard 6, 8000 Aarhus C, DK, Kaj S. Christensen
Send response to journal:
|
We have concerns about the conclusions of Jellema and colleagues.1 1) The authors made two interventions. GPs in the minimal intervention strategy group were a) trained to make psychological assessments, and b) asked not to refer patients to a physiotherapist in the first six weeks. According to the paper, more patients in the usual care group than in the minimal intervention strategy group visited a physiotherapist (39 vs. 18%). The authors’ conclusion that general practitioners should not adopt a treatment strategy aimed at psychosocial prognostic factors is therefore not supported by their findings. The fact that they found no effect of the minimal intervention strategy may actually indicate that psychological assessment is effective in reducing the use of physiotherapy without any negative effects on patient health. 2) The authors do not consider) whether the lack of effect may be explained by methodological flaw. They do not mention two major methodological shortcomings tending to bias the result toward no effect. The first concerns the way patients were sampled and psychologically characterised. The GPs were asked to include 10 consecutive patients with low back pain (LBP). The inclusion criteria were broad and at least half of the patients had experienced an episode of LBP in the previous year. For 20% the episode was characterised as an exacerbation. We may thus be dealing with a more chronic condition although the authors wanted to study acute LBP. Were the patients selected by the GPs as "good patients" (only 5.2 patients per GP were actually included)? The authors reject the possibility of selection bias because the patients were similar at baseline. But this may also prove that selection bias was present in both randomisation groups. The authors do not mention that some of the patients may be misclassified in that they did not suffer from LBP. This may bias effect measures toward the 0-hypothesis. Unfortunately, the authors do not report whether internal validity has been assessed. Psychological distress was assessed by a research assistant without feedback to the GPs. But how were the patients then characterised in a more clinically relevant way? GPs may not be able to integrate a 4DSQ score of 11 in clinical practice! Did the patients show any signs of depression, anxiety or, perhaps most relevant, somatisation? Despite the randomised design, we do not know the number of clinically relevant psychosocial symptoms and health problems. In theory, there may not have been any! 3) It appears from the text that the authors used parts of larger validated questionnaires. It is generally recognized among questionnaire users that one cannot select sections of a questionnaire and expect the new questionnaire to have the same validity as the original. Moreover, the authors used a statistical instead of a clinical cut-off. We fear that the characterisation of the patients, especially in terms of psychological measures, was insufficient. This may lead to misclassification of patients and thus bias the effect towards no result. 4) The authors state that they performed “intention to treat” analyses. However, some patients withdrew from the study during the study period and were not included in the analyses. We are also concerned about the use of odds ratio analyses in a randomised intervention study (person time) in a high prevalence population (>20%). However, the choice of method may be due to methodological and/or computational considerations, which the authors unfortunately do not or discuss. 1. Jellema P, van der Windt DAWM, van der Horst HE, Twisk JWR, Stalman WAB, Bouter LM. Should treatment of (sub)acute low back pain be aimed at psychosocial prognostic factors? Cluster randomised clinical trial in general practice. BMJ 2005;331:doi = 10.1136/bmj.38495.686736.E0. Competing interests: None declared |
|||