Age and Ageing Advance Access originally published online on January 18, 2006
Age and Ageing 2006 35(2):185-187; doi:10.1093/ageing/afj028
© The Author 2006. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org
Senior physicians’ estimates of the likely effects of feeding policies on outcomes prior to the completion of the FOOD trials
SIR—The FOOD trials were a family of three multicentre
international randomised controlled trials that aimed to estimate
the effects of different feeding policies on outcomes of stroke
patients [
1,
2]. Their purpose was to identify the most effective
feeding interventions, which could be included in guidelines
so that variation in practice could be reduced and patient outcomes
improved. In this study, we aimed to assess the likely impact
the trial results would have on actual feeding practices by
comparing physicians’ estimates of treatment effects with
those from the trials themselves.
 |
Methods
|
|---|
Prior to trial completion, we surveyed physicians to establish
their estimates of the direction and size of treatment effects.
Between November 2003 and February 2004, we sent a questionnaire
to 334 consultant physicians in the UK who regularly managed
stroke patients. We asked them to estimate the absolute differences
in the proportions of patients receiving our different feeding
allocations who would have a poor outcome [death or modified
Rankin scale (MRS) 3–5] in FOOD Trial 1 (oral supplements
versus avoid supplements) and death or MRS 4–5 in Trial
2 (early versus avoid tube) and Trial 3 [percutaneous endoscopic
gastrostomy (PEG) versus nasogastric tube (NG)]. We provided
an estimate of the proportions of patients with a poor outcome
across both treatment groups in each trial on the basis of an
interim analysis of our data performed in October 2003. To aid
clinicians who were unfamiliar with expressing treatment effects
in terms of absolute benefit, we provided two examples based
on familiar stroke treatments, aspirin and rtPA. Non-responders
were sent at least one further questionnaire, supplemented by
personal approaches (in person and by telephone) by the principal
investigator to recipients known to him. Data were entered into
an Excel spreadsheet. The distributions of estimates of effect
size were plotted for each of the trials. Median and interquartile
ranges were used to describe the distribution of estimates of
effect.
|
Results
|
|---|
We received 206 (62%) completed questionnaires. The distributions
of clinicians’ estimates of effect are shown in Figure
1.
In Trial 1, 179 (90%) clinicians’ believed that oral supplements would improve patients’ outcomes and 53% provided estimates within the 95% confidence intervals (CI) of the trial estimate. Forty per cent believed supplements would increase the absolute proportion with a good outcome by at least 3%—this size of benefit is statistically very unlikely according to our trial results [1].
In Trial 2, 169 (85%) thought that early tube feeding would provide better outcomes and 67% of clinicians’ estimates lay within the 95% CI of the trial estimate. The trial’s estimate with respect to death or MRS 4–5 was only marginally in favour of early tube feeding although the estimate for death alone was a –5.8% (95% CI –12.5 to 0.8%) absolute reduction [2]. This estimate would be more in line with clinicians’ expectations.
In Trial 3, 158 (79%) clinicians expected PEG feeding to improve patients’ outcomes compared with NG feeding. This was opposite to the effect we actually observed. Despite being the smallest trial with the widest 95% CI, only 21% of clinicians’ estimates lay within them. Of course, our trial result with respect to death alone, where we saw only a small decrease with NG feeding (1.0% (95% CI –10.0 to 11.9)), was more in keeping with clinicians’ views [2].
|
Discussion
|
|---|
There have been rare attempts to directly measure the impact
of a multicentre randomised trial on clinical practice by surveying
practice before and after the trial. This approach is not possible
here because we have no reliable surveys of actual practice
before the FOOD trials were published. Also, in some trials,
the investigators have tried to establish the prior beliefs
of clinicians, usually in small numbers of experts, to enable
them to incorporate a Bayesian approach to trial monitoring
[
3]. Our data represent the views of a large number of physicians,
but the response rate was moderate so they may not be truly
representative of all UK physicians. The responses showed some
digit preference for 0, 1, 5 and 10% differences. Some estimated
absolute differences of 10% or more which reflects a tendency
for clinicians to over-estimate the likely treatment effects
and contributes to the type 2 errors observed in many randomised
trials. Although we asked them to estimate the effects on ‘death
or poor outcome’, it is unclear whether their estimates
would have differed substantially for the effect on death alone.
This may be important because in Trial 2 and Trial 3 we saw
a marked disparity, at least in our point estimates, with respect
to the different outcomes.
Our simple approach provides a practical and relatively inexpensive method to assess the likely impact of large trials and systematic reviews. We would argue that if a trial’s results are compatible with clinicians’ estimates, then the trial is likely to reinforce, rather than substantially change, clinical practice. However, where the clinicians’ estimates differ in direction of observed treatment effect, as in Trial 3, then one might expect a major change in practice.
|
Key points
|
|---|
- Until the FOOD trials reported their results in February 2005, no reliable evidence existed regarding the best methods of feeding stroke patients whilst in hospital.
- This survey shows that most UK physicians believed, prior to dissemination the FOOD trial results, that:
- oral nutritional supplements and early enteral tube feeding would improve patient outcomes—results which are broadly in line with the trial results.
- PEG tubes would result in better outcomes than NG tubes which the trial indicated was very unlikely to be the case.
- The FOOD trial results are likely to reduce the use of PEG feeding early after stroke.
Martin S. Dennis*,
Steff Lewis and
Gina M. Cranswick
Clinical Neurosciences, University of Edinburgh, Edinburgh, Scotland, UK Email: martin.dennis{at}ed.ac.uk
* To whom correspondence should be addressed
|
Acknowledgements
|
|---|
The FOOD Trial was supported with grants from: the Health Technology
Assessment Board of NHS Research and Development in UK, the
Stroke Association, the Chief Scientist Office of the Scottish
Executive and Chest, Heart and Stroke Scotland. The views expressed
in this paper are those of the authors and do not necessarily
represent those of the funding bodies. The funding bodies played
no part in the design, data collection, analysis or drafting
of the manuscript. We thank Janie Hunter for her help in sending
out questionnaires, entering data and managing the database.
|
References
|
|---|
- The FOOD Trial Collaboration. Routine oral nutritional supplementation for stroke patients in hospital (FOOD): a multicentre randomised controlled trial. Lancet 2005; 365: 755–63.[Medline]
- The FOOD Trial Collaboration. Effect of timing and method of enteral tube feeding for dysphagic stroke patients (FOOD): a multicentre randomised controlled trial. Lancet 2005; 365: 764–72.[ISI][Medline]
- Spiegelhalter DJ, Abrams KR, Myles JP. Prior Distributions in Bayesian Approaches to Clinical Trials and Health-Care Interventions. Chichester: John Wiley & Sons Ltd, 2004, 139–80.
CiteULike 
Connotea 
Del.icio.us What's this?
Top
Methods
Results
