Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Commentary on the paper by Balfour-Lynn et al
In a retrospective review published in this issue,1 Balfour-Lynn and colleagues describe 16 children with cystic fibrosis (CF) who appeared to show clinical improvement following regular infusions of intravenous immunoglobulin (IVIG). They have not described any criteria for the commencement of treatment, but the majority of children had previously been diagnosed with allergic bronchopulmonary aspergillosis (ABPA). An analysis of efficacy compared lung function and other concomitant treatments before starting therapy and after courses of therapy, the number of which varied considerably between patients. This treatment was associated with a reduction in the doses of oral and inhaled steroids. There was some improvement in forced vital capacity, but no difference in forced expiratory volume in one second.
Clinical practice has been likened to an experiment, where a patient presents with a problem, treatment is initiated, and the results of treatment are later assessed and conclusions drawn about whether or not the treatment is effective. There are a number of factors which may act to bias these conclusions.2 Firstly the patient may have improved anyway. Secondly, new treatments are more likely to be initiated at a time when, as part of the normal fluctuation in the illness, symptoms, signs, and other measures have reached temporary extreme values which, will “regress towards the mean” when next measured. Any treatment started between the time points will appear to be effective. Thirdly, there is a well described placebo effect and finally, in the absence of blinding, the expectations of the patients and/or clinicians may bias the assessment of whether or not the treatment has been effective. This is particularly the case when outcome measures are based on clinical decisions, such as use of other concomitant therapies.
The n of 1 trial, in which randomised, double blind, multiple crossover comparisons of active and placebo treatments are conducted in an individual patient,2 has been developed to limit the biases inherent in uncontrolled therapeutic trials. There are a number of situations in which n of 1 trials may be considered. For example, as proposed by these authors, there are situations where clinical trials have not been conducted or may not be feasible.2 It has also been suggested that they can be used in the early phases of drug development, to assist in the planning of subsequent large clinical trials.3 Finally, where a treatment has been shown in clinical trials to have an overall benefit, but there is considerable individual variation in response, n of 1 trials can identify which patients are likely to derive benefit. To address the last of these questions, n of 1 trials of recombinant DNase have been successfully performed in CF patients.4,5 Prerequisites for conducting n of 1 trials are that the underlying condition should be stable over time, the treatment should exert its effect over a short time, and this effect should stop quickly once the treatment is withdrawn.6
Uncontrolled clinical trials, such as those described in the paper by Balfour-Lynn and colleagues,1 have important limitations. Is the alternative, suggested by these authors, of undertaking n of 1 trials of intravenous immunoglobulin in “carefully selected patients” with CF, appropriate? It is not clear from this report whether the patients with severe CF lung disease, who were included, represent a group whose lung disease would be stable, at least over the duration of an n of 1 trial. It is implied in the report, that the response to treatment may be cumulative and therefore multiple crossovers of one month’s active treatment and one month’s placebo treatment may not enable an adequate assessment. Similarly the observation that improvement was maintained after stopping treatment or indeed that the illness was “switched off” completely suggest that this therapy is not amenable to assessment by n of 1 trials. As others have observed,4 setting up and complying with the protocol in an n of 1 trial requires considerable time and commitment on behalf of patients and clinicians. The authors of this report have suggested that a randomised controlled trial would not be ethical because of the need to administer placebo in the control group, but this would also the case in an n of 1 trial.
What then is the appropriate way to assess effectiveness of treatments in a small, heterogeneous group of patients such as this? These authors have alluded to the problems of conducting clinical trials in a rare disease such as CF, particularly where only a small subset of patients would be eligible. The majority of this group had a previous diagnosis of ABPA and in a proportion it was judged to be still active at the initiation of IVIG. It is possible that the clinical benefits were due to a modulation of this complication. ABPA is being increasingly recognised in CF and a search of The Cochrane Library7 has revealed no clinical trials of any therapy for this condition in CF. It is likely that there are sufficient CF patients with ABPA nationally to provide an adequate sample size for a UK-wide clinical trial of such therapies, and this could be facilitated by assessing numbers from the UK Cystic Fibrosis database.
I would suggest that the way to approach evaluation of treatment in such challenging clinical settings is to first consider randomised controlled trials. The exponential increase of randomised controlled trials in CF in the past 30 years suggests that some of the feasibility issues are being overcome.8 This may involve a certain amount of “thinking outside the box” and considering carefully what the components of each complex clinical problem might be. N of 1 trials may certainly have a place, but careful thought should be given to whether the condition and the treatment are amenable to assessment in this way. Observational studies may be helpful, particularly in the assessment of adverse effects.9 Case reports such as this1 are valuable, but only insofar as they generate hypotheses which lead to definitive studies rather than on their own providing conclusive evidence on which to base treatment.
Commentary on the paper by Balfour-Lynn et al