Source: Bandolier
Clinical bottom line
As well as well understood biases, new forms of bias or potential bias in chronic pain studies are emerging. Unless we take care, we can make the wrong decisions when comparing therapy efficacy unless like is compared with like, and at the highest level of evidence.Moore et al. "Evidence" in chronic pain--establishing best practice in the reporting of systematic reviews. Pain 2010 150: 386-389.
Starting point
We know that there are limitations in evidence, and that higher quality studies produce, almost overwhelmingly, more conservative results than those with less rigorous methods. Understanding exactly what quality, or validity, means, is a dynamic, and we have to keep relearning lessons and ratcheting up the minimum standards we accept for "evidence". This review casts a cold and fishy eye over evidence in chronic pain trials.Suggestions
The review makes the following points about what constitutes good evidence:- Randomisation, preferably properly done and concealed.
- Blinding, preferably properly done so that all concerned are unaware of treatment.
- Imputation method. This is a new one, and recognises that withdrawal rates in chronic pain trials can be high - up to 50-60% in chronic low back pain. Right now, the last observation carried forward method is commonly used, so that patients not taking the medicine can contribute to efficacy estimates. The recommendation is that this not be done, and that only those patients taking the medicine contribute, as no analgesia can come if you don't take the medicine. Right now there is little evidence on this, but more is emerging, and this could have a major effect on efficacy estimates.
- Study duration should be reasonably long - ideally 12 weeks. There is increasing evidence from longitudinal individual patient analyses that short studies can overestimate treatment effects, particularly for less effective therapies.
- Average pain score data will mislead. Responder analyses should be used instead. Again, individual patient data analysis indicates that distributions in pain trials are U-shaped, not Gaussian, making average values rather silly.
- There may be exaggerated treatment effects in crossover trials. and a predominance of crossover trials may be considered a possible source of additional bias.
- Size is crucially important, and when there are fewer than 200 events (an event would be a patient achieving a given level of response, for example), then results cannot be trusted, simply because of the magnitude of random chance effects.
Core outcomes
It seems likely that future systematic reviews, and trial analysis, will be based on some core outcomes of benefit and harm. A likely set of outcomes for chronic pain is likely to include some or most of the following:- Pain
- - At least 50% pain reduction
- - At least 30% pain reduction
- - Proportion below 30/100 mm (no worse than mild pain)
- - Patient global impression (very much improved)
- Function
- General
- - Quality of life measure
- - Patient global impression
- Adverse events
- - Withdrawal due to adverse event
- - Serious adverse events
- - Death
Making comparisons
Indirect comparisons between treatments can be made where there is an adequate amount of good quality data. Comparability is imperilled when like is not compared with like - when, for example, results from small, short studies with easily-attained but inadequate outcomes are compared with large, long duration studies with clinically relevant outcomes that are hard to attain. It is also imperilled when different doses or treatments are erroneously combined under a general label and then discussed as identical. Meaningful comparison of efficacy with other interventions is not possible where quality, validity, and size standards are not met by any one of the comparators.Yet making comparisons between treatments is what we try to do all the time, in making decisions about individual patients, when making policy, and making guidelines. The bottom line is that we will have to be much more careful in future.
No hay comentarios.:
Publicar un comentario
Write here your comment