Are clinical trials relevant?

Last month, an interesting article in Transplantation by Blosser et al. (Transplantation 91(8):858) explored the issue of generalisability of clinical trials to patients outside of the clinical trial setting.  The authors identified 573 randomised clinical trials (RCTs) in kidney transplant recipients in the period between 1999 and 2008, and found that 30% had an exclusion criteria based upon age, with 16% excluding the over 65s.  Given that the average age of renal transplant recipients is increasing over time, and that 16% recipients in the US in 2008 were over 65 years of age, this does raise some concern as to whether the results of such trials can be generalised to older recipients.  Comparing data from these trials to the US OPTN database, Blosser and colleagues found that the average trial recipient was 5 years younger than the entire cohort of recipients in the same time period (45 years vs. 55 years).

Such concerns about the artificial patient population created in the clinical trial setting have been recognised in the past as a limitation of randomised controlled trials.  Trials often take place in teaching or university hospitals, with more intensive monitoring of clinical conditions, greater funding and resources and perhaps a lower threshold for intervention than a non-trial setting.  Thus whether the effect sizes seen in such trials would be as great in a more “normal” district general hospital population is often debated.  It is very likely that the effects seen in some trials will at the very least be reduced in the “less perfect” non-trial environment.

Of course, such concerns are probably less in the transplant population as most transplant activity takes place in larger teaching centres even outside of the trial setting.  What is of more concern, as highlighted in the paper from Blosser and colleagues, are inclusion/exclusion criteria to trials that create an artificial population often very different from our overall transplant cohort.  As well as age, these criteria may include immunological risk factors and comorbidities.  The effects of treatment in such healthy, young and low-risk patients may not be as dramatic or may differ entirely in an older recipient with a previous transplant, diabetes and ischaemic heart disease!

How can we deal with these discrepancies in patient populations?  Multi-centre trials and systematic review across a number of trials may help to iron out some of the centre-specific effects, but more important is consideration of the target population at the design stage of a trial.  If the population of interest is the general transplant population in a unit, then every endeavour must be made to include a representative cross section of patients in the trial irrespective of age, comorbidity or immunological risk.  Obviously safety and ethical issues must be taken into account.  This may be facilitated my more attention to the issue from funding bodies and journal editors who should review and question inclusion and exclusion criteria if they appear to be too restrictive.

An aside to all of this is the role of observational data.  We are lucky in the transplant world to have access to very complete and comprehensive clinical registry data, and one of the oft-cited advantages of such data is that it reflects the “real world” more accurately than clinical trial data.  If the inclusion of data in a registry is mandatory, as the the UK and the US, then examination of outcomes across a larger number of more representative patients is possible.  Making treatment comparisons using such data is, however, very difficult as there is no control over recruitment to the treatment groups meaning a significant risk of selection/allocation bias.