Abstract
Background
Up-to-date summaries of current research evidence are needed to inform clinical decision-making. A variety of methodologies is now available for synthesising the data from different primary intervention studies including traditional meta-analysis, IPD meta-analysis and mixed-treatment comparison
meta-analysis. Our objectives were to identify areas of high decision uncertainty in collaboration with our NHS colleagues and summarise the best available evidence using appropriate techniques. Methods
Methods included stakeholder consultation to identify decision uncertainties; a scoping review of the evidence for silver-containing wound dressings for treating venous leg ulcers; application of Cochrane methods of systematic review to 11 complex wound topics; and mixed-treatment comparison meta- analyses of dressings for diabetic foot ulcers and venous leg ulcers.
Results
Techniques involving facilitated, face-to-face contact with health professionals performed best in generating clinical uncertainties as topics for evidence synthesis. The relative effectiveness of different wound dressings for different wound types had high priority. There was no evidence that silver dressings were more effective than non-antimicrobial dressings for healing venous ulcers; however, variability in the existing trials precluded IPD meta-analysis. A series of Cochrane reviews in prioritised topics identified several wound treatments that appear to be more effective than others in different wound types. The matrix hydrocolloid dressing was associated with the highest probability (70%) of being the best dressing for diabetic foot ulcers, whereas a hyaluronan fleece dressing had the highest probability (35%) of being the best for venous ulcers; however, in both cases there was high uncertainty and poor-quality evidence. Conclusions
A range of approaches to evidence synthesis was applied to complex wound treatments across a broad range of topics that had been prioritised by health-care professionals. This approach identified some treatments associated with the highest probability of effectiveness.
Background
To date, the Cochrane Wounds Group has identified around 9500 clinical trials in wound management and assembled them on its Specialised Wounds Register [see http://wounds.cochrane.org/ (accessed 11 May 2016)]. Evidence synthesis methods, such as systematic review, enable the management of information from multiple primary studies by identifying, summarising, appraising and pooling groups of primary studies in relation to a predefined research question.183The standard systematic review approach detailed in theCochrane Handbook for Systematic Reviews of Interventions163describes methods to manage information, reduce subjectivity, establish generalisability and consistency, improve power and precision and identify gaps in the research. The aim is usually to derive a summary estimate of effect from multiple primary studies (e.g. in terms of treatment effectiveness) while minimising bias in the review process. To date, systematic reviews conducted under the auspices of the Cochrane Wounds Group have contributed important information to the evidence base in wound management. For example, we know
that compression is an effective intervention for healing venous leg ulcers184and that higher-specification foam mattresses reduce the incidence of pressure ulceration compared with standard hospital foam mattresses.180This said, it should be acknowledged that the standard systematic review methods can be restrictive as they are based mainly on pairwise comparisons and group-level data. Other, more advanced, systematic review and meta-analysis methods are now available that may be more informative in areas of persisting uncertainty. These include those based on IPD and those involving mixed-treatment comparison meta-analysis.
Individual patient data meta-analysis involves the identification and retrieval of original patient-level data from the primary investigators, who are invited to become research collaborators along with the review team. Data are recoded, checked, cleaned and reanalysed. Advantages include the potential to conduct powerful time-to-event analyses adjusted for predictive patient-level covariates. This helps to obtain a more precise estimate of treatment effect and so reduce uncertainty. Other advantages include opportunities to conduct meaningful subgroup analyses based on patient-level factors, reinstate patients who the primary investigators have excluded from their own analyses, include updated data on the event in question from follow-up beyond the trial end point and combat poor reporting through close collaboration with the primary investigators.185As such, IPD meta-analysis has been proposed as the‘gold standard’among systematic reviews because of its potential to generate estimates on the most complete and clean data set possible with full adjustment for predictive covariates.186It is particularly useful when there is considerable uncertainty (imprecision) despite the existence of a good-quality systematic review of trial-level data, together with a suspicion that important patient-level covariates are contributing to the imprecision. One such example from wound care relates to the comparison of the effects of the four-layer compression bandage and the short-stretch bandage on the healing of venous leg ulcers. In this case, an earlier
Cochrane review had not shown a difference in healing between the two bandages when trial-level data were pooled,184whereas a subsequent IPD meta-analysis of the same comparison, involving estimation from an adjusted Cox proportional hazards model, showed that, on average, patients healed faster with the four-layer bandage.187The main issue with undertaking IPD meta-analysis is that the success of the endeavour is contingent on the review team being able to access the original data for each eligible primary study, although IPD meta-analysis may still be beneficial even if data can be obtained for only a subset of eligible studies. Access to the original patient-level data may not always be straightforward; for example, data may no longer be available for older studies. In addition, such reviews take longer and are more resource intensive than standard systematic reviews of group-level data.
Pairwise comparisons of several types of interventions within a review can provide some useful information, depending on the number of different interventions available. However, when there is a range of competing interventions, multiple pairwise comparisons may not be informative for decision- makers, particularly when active treatments are always compared with placebo or usual care. In such cases it can be impossible to draw valid conclusions regarding the relative effects of several competing
technologies and standard meta-analysis does not lend itself to the ranking of treatments in terms of their estimated effectiveness. Wound dressing selection is an aspect of health care, like many others, in which there are many competing interventions for the same indication, in which strong claims are made in the marketing literature regarding relative treatment effects and in which clinicians understandably struggle to make informed choices.188Mixed-treatment comparison systematic review and meta-analysis enables more than two different interventions to be compared simultaneously by extending the standard method of meta-analysis using a Bayesian model to utilise both direct and indirect comparisons within a network of evidence. Mixed-treatment comparison meta-analysis also provides the opportunity to rank interventions according to their probability of being the best intervention relative to the others in the network.189Issues with the mixed-treatment comparison approach include the need to meet assumptions such as consistency between evidence derived from direct and indirect comparisons190and how to deal with variation in risk of bias across different parts of the network.191As with standard meta-analysis, variation in risk of bias across
With a choice of possible review methods in mind, workstream 3 began with a consultation exercise with our NHS partners to identify high-priority topics for evidence synthesis. This was followed by consideration of suitable topics for advanced methods of evidence synthesis, as well as earmarking those appropriate for new and updated Cochrane reviews. The aims of this were to provide some immediate evidence to inform practice by making maximum use of existing evidence as well as to identify gaps in the evidence to inform future research. In this chapter we report the methods and results of the following four pieces of work: 1. the identification and prioritisation of topics for evidence synthesis
2. selecting candidate topics for IPD meta-analysis
3. undertaking new and updating existing high-priority Cochrane systematic reviews 4. mixed-treatment comparison meta-analysis in high-priority topic areas.