Phase III Randomized Controlled Trial of a Breathlessness Intervention Service for Intractable Breathlessness.
1. Is BIS more effective than standard care for patients with intractable breathlessness
from advanced malignant or non-malignant disease?
2. Does it reduce patient & carer distress due to breathlessness, & increase patients'
sense of mastery of the symptom?
3. What are the experiences & views of those who use BIS, their informal carers & the
clinicians who refer to it?
4. Is BIS cost-effective?
To assess changes in patient outcomes attributable to BIS, a pragmatic RCT design will be
conducted using a fast-track design (design proved highly acceptable at Phase II). Analysis
will focus on comparing patients who have used BIS to those who have not (yet) used BIS (the
use of BIS by the control group will occur subsequent to these measurement points).
The researcher will be blinded to the allocation of respondents up until discharge/referral
on from BIS for FT patients or at the end of the waiting period for CC patients. This will
be achieved by the researcher conducting study recruitment & collecting baseline measures
before instructing a third party (local clinical trials' nurse) to conduct randomisation &
reporting of allocation (to the patient & BIS) using a random sequence of opaque envelopes
previously generated by a statistician independent of BIS. Subsequently, all data will be
handled using study identity numbers; group allocation identifiers will only be added at
analysis. This model was used successfully in the Phase II pilot trial.
Data collected from the control group once they are in receipt of BIS (after their period on
the waiting list when their group allocation was blinded to the researcher) will be treated
as before/after data & not RCT data. This will allow the collection of qualitative data from
this group at the midpoint of using BIS.
The two broad disease courses, malignant (m) & non-malignant (nm), will be considered
separately due to their different trajectories & needs, & resultant different service
models. The intervention model for patients with non-malignant disease consists of two-three
visits & three telephone contacts (to patient &/or primary care staff) over a four-week
period with a 16 week (from first assessment) follow up, whereas the model for patients with
malignant disease consists of one visit in conjunction with a primary care professional/key
worker & two telephone contacts (to patient &/or primary care staff/key worker) over a
two-week period, with a six week (from first assessment) follow up . Thus, the measurement
points for the disease groups will differ.
Data (quantitative/qualitative) will be collected for all respondents at baseline (t1),
prior to randomisation. For those with non-malignant conditions (nm), this will be repeated
for the FT group (nmFT) midway through the intervention (two weeks post commencing BIS;
nmFTt2 - quantitative data only) & at the equivalent time point for the CC group (i.e. two
weeks from entering the waiting list; nmCCt2 - quantitative data only), then again after
discharge/referral on from BIS for the fast track group (four weeks post commencing BIS;
nmFTt3) & four weeks after discharge (nmFTt5). As well as being interviewed at t1
(randomisation) & t2 (midway between randomisation & BIS - quantitative data only), the CC
group (nmCC) will also be interviewed just prior to commencing BIS (nmCCt3), during BIS (two
weeks post commencing BIS; nmCCt4) & then again after discharge/referral on from BIS (four
weeks post commencing BIS (nmCCt5). This will allow identification of whether or not
respondents in the control group deteriorated significantly whilst waiting for BIS & the
impact of this on final outcomes. This model was successfully piloted at Phase II.
For those with malignant conditions (m), baseline measures (t1) will be repeated for the FT
group (mFT) after discharge/referral on from BIS (two weeks post commencement of BIS; mFTt3)
& two weeks after discharge (mFTt5). For the CC group (mCC) baseline measures (t1) will be
repeated just prior to commencing BIS (mCCt3) & then again after discharge/referral on from
BIS (two weeks post commencing BIS; mCCt5). Thus, due to the shorter time frame of the
malignant service model, there would be no t2 or t4 measurement points.
Sample size is based on the existing literature & on our experiences at Phase II. The
estimated standard deviation of the primary outcome measure is 2.5. In order to detect a
2-point difference in mean outcome between groups (equivalent to 0.8sd effect size) with 80%
power using a t-test at the 5% level of significance, it will be necessary to recruit 26
patients per arm per disease group (trial) followed up to provide an outcome, which will now
be at an earlier time point to minimize attrition. By adjusting for the baseline of the
primary outcome measure in the analysis using analysis of covariance we anticipate an
improvement in the precision of the estimated intervention effect. We also anticipate an
improvement from the use in the trial of the NRS measure rather than the more highly
variable VAS version from the pilot used in this sample size calculation. Therefore we
propose to recruit 120 patients, 60 per disease group to ensure adequate power, effect size
and allowance for attrition.
Analysis will be on an intention to treat basis. Primary analysis will be by analysis of
covariance of the NRS at the final time-point with adjustment for baseline NRS in order to
improve precision of the estimated BIS versus comparison effect. Secondary analysis will
more sensitively incorporate all time-points in a repeated measures linear regression to be
able to detect any differences between BIS & comparison emerging/changing over time within
the initial period. All analysis will be documented prior to final data collection in an
analysis plan, will be addressed with two-tailed tests & be assessed at the 5% level of
significance. Effects within & between arms will be summarized using means & 95% confidence
intervals. If randomised effects are similar in malignant & non-malignant groups, a combined
effect will be estimated allowing a narrower overall confidence interval. Analysis will be
conducted using SPSS software.
The cost of BIS for each patient will be calculated from service activity data combined with
staff salary costs, plus on-costs, overheads & equipment. Information on the use of other
health & social services & informal care (proxy-valued as a homecare worker) will be
collected with the Client Service Receipt Inventory. This will be combined with appropriate
unit cost data (Curtis & Netten, 2006) to generate service costs. Cost comparisons will be
made using bootstrapping methods to account for any skewness in data distribution.
Cost-effectiveness will be assessed by combining cost data with that on outcomes including
quality-adjusted life years (QALYS), in the form of incremental cost-effectiveness ratio &
acceptability curves. Qualitative data will be analysed using a framework approach (Ritchie
& Spencer, 1993) conducted using NVivo software.
Allocation: Randomized, Endpoint Classification: Efficacy Study, Intervention Model: Parallel Assignment, Masking: Single Blind (Investigator), Primary Purpose: Supportive Care
Numerical rating Scale (NRS) for distress due to breathlessness
End of intervention (4 weeks after baseline for patients with a non-malignant diagnosis; 2 weeks after baseline for patients with malignant diagnoses)
Sara Booth, FRCP
Cambridge University Hospitals NHS Foundation Trust
United Kingdom: Research Ethics Committee