1. El hecho jurídico
2.10 Diferencias normativas
I aim to describe consecutive the steps taken, in order to develop a new methodological framework for trial eligibility assessment. The framework is designed to be easily understood by clinicians and implemented in real time during the course of a trial. This process brought together a number of research methods across different disciplines. Their complementary and complex relationship is illustrated in Figure 3.1, accompanied by an explanation in the text.
The literature review in chapter 1.2 indicates that Clinical Equipoise as the basis for setting up a clinical trial is usually demonstrated by a failure of consensus between experts about the effectiveness of an experimental intervention compared to a control. This is usually caused by the lack of knowledge that enables experts to forecast the outcome of a proposed intervention. A “measure of a state of knowledge” (Jaynes and Bretthorst 2003) can be demonstrated using the concept of Bayesian subjective probability. Bayesian methods allow one to quantify the level of an
expert’s individual uncertainty or confidence about the effectiveness of an intervention. When experts’ opinions are combined, or pooled, it is safe to assume that for most patients in a trial population, the experts should fail to reach an agreement in presence of clinical equipoise, which is necessary for a patient to be recruited ethically. This concept provides a platform from which to develop ideas in the challenging trial setting.
An expert treatment prognosis can be viewed as a Bayesian prior which is assigned to a specific hypothesis; it is personal and varies with an individual’s knowledge and expertise. However, turning informally expressed ideas into a mathematical prior distribution is perhaps one of the most difficult aspects of Bayesian analysis (Spiegelhalter, Abrams et al. 2004). There are five widely used approaches: (i) elicitation of subjective opinion, (ii) summarising past evidence, (iii) default priors, (iv) ‘robust’ priors and (v) estimation of priors using hierarchical models. There is no such thing as a correct prior or method of determination, but option (i)
is the most suitable to elicit and quantify the collective subjective opinion from a panel of experts in real time. It has the advantage of being dynamic and flexible, because knowledge and preferences can change during
the course of a trial. This may happen, for example, on the publication of related research, or as individual and collective clinical experience accumulates amongst experts.
Development of an online tool for opinion elicitation has been the first step to bring the Clinical Equipoise assessment in real time. Previously, elicitation of subjective opinion in surgical settings has been achieved by collecting opinions through a questionnaire survey (Young, Harrison et al. 2004) or by a series of scenario-based specialist interviews (Lilford 1994). Both methods are time and labour intensive and could only be used to support a justification to start clinical research. An alternative technique, which involves the collection of subjective judgements in relation to clinical decision making, allows participants to distribute 100 points between bins that reflect their ‘weight of belief ’ in a range of available outcome options (Parmar, Spiegelhalter et al. 1994; Parmar, Griffiths et al. 2001). This technique was used to develop a novel web-based tool for the collection and measurement of specialist beliefs about a specific clinical case (Chapter
4.1). The technical implementation was achieved in collaboration with Mr S
Brydges, eLab, the University of Warwick. Freely available software (Adobe Flash Player, MSDN Microsoft Data tables) and the University of Warwick web-based platforms (Warwick Forums, SiteBuilder) were used. The
tool was then tested in a pilot study. A questionnaire, as well as informal feedback, was used to assess expert perception and usability of the tool. I was interested in technical issues, format, clinical information sufficiency and ultimately whether I had the experts’ support to take the study further. Accordingly, a combination of closed (choice of categories) and open questions was used to guide responses (Appendix K). The questionnaire was designed in an easy to read one page format with spaces for comments (Boynton and Greenhalgh 2004). It was posted to surgeons’ secretaries upon completion of the pilot study.
Once the voting data for a case are obtained, two potential approaches to pooling the opinions are available, (i) a parametric model based on a Beta distribution and (ii) a nonparametric model based on estimated means and standard deviations. Statistical models of expert opinions were developed in collaboration with Dr N Parsons, Medical Research Statistician, Statistics and Epidemiology, Warwick Medical School. An
Opinion Elicitation Course was attended (CRiSM: Centre for Research
in Statistical Methodology, 13/04/2011, University of Warwick).
Subjective logic principles (Jøsang 1997) were used to develop a statistical model that allowed us to measure and visually present a level of clinical equipoise for an individual clinical case. The formal statistical methods necessary to implement the model, which are outside the scope of this thesis, are described in detail elsewhere (Parsons, Kulikov et al. 2011) (see
and interpretation of clinical equipoise levels (Chapter 4.2). In particular, decisions about trial eligibility depend on the application of decision rules. There is little research available on this subject, other than a stand-alone ground-breaking paper where collective equipoise levels sufficient for initiating a clinical trial were estimated (Johnson, Lilford et al. 1991). This ethometric study with the general public investigated how much collective equipoise can be disturbed before potential trial subjects deem it to be unethical. Their findings suggested that “trials are perceived as unethical when equipoise is disturbed beyond 70:30. In other words, when 70 per cent of experts favour treatment A, then 50 per cent of subjects would prefer that treatment A be administered rather than subjected to critical assessment.” When 80% of experts favour one treatment, less than 3% of the lay public would consider human trials morally justifiable. Based on these estimates, decision rules were suggested and this complemented the development of the Patient Eligibility Assessment through Clinical Equipoise (PEACE) methodological framework.
To test the new methodology, the PEACE framework was introduced as an independent research project called ‘Collective Uncertainty Project’ within the UK HeFT (Chapter 4.3). The aim was for real clinical cases to be assessed in real time by an expert panel consisting of principal investigators in the context of a challenging trauma trial, but without interfering with the clinical management or the trial course itself. This was achieved by asking eligible patients to consent to the use of their clinical data at least 6
weeks after the injury, when a decision about trial participation was made and a treatment course was initiated. The use of ethical approval was kept separate from the main trial (Appendix C). This had the additional advantage of approaching both patients who agreed to take part in the UK HeFT and those who did not. Surgeons who were involved in case assessments as part of the expert panel were asked for their feedback
informally and via questionnaires (4.3.1). On this occasion the questionnaire design was more complex (Appendix G). Although anonymous, some background information was requested from experts to reflect levels of specialist surgical and research experience. Data examples were given and the choice of categories included an open element to stimulate reflection of the study involvement experience.
Finally, patient understanding and views about the possible introduction of the new framework in future trials were researched, as part of the qualitative study (using semi-structured interviews), which is described later in this chapter (3.3).
3.2 Methodological improvement to patient recruitment