Clinical research

Research setting

Medical research is conducted across a number of different settings. Different components flow, in two directions, with clinical problems and observations being investigated at a fundamental level and improved fundamental knowledge being applied to the patient setting. The resultant interplay of ideas, questions, solutions and new questions ( Fig. 37.1 ) is what defines the process of medical research. Whilst the boundaries can merge, broadly the style of research being conducted allows these different settings to be considered separately.

Fundamental (or basic) research seeks to develop our understanding and knowledge of the genetic, molecular, environmental or societal basis of disease. Examples include:

• Investigating the cardiovascular or central nervous system effects of neonatal asphyxia using an animal model

• Identifying the cytogenetic abnormalities found in acute lymphoblastic leukaemia

• Identifying the underlying cause of the abnormally thick secretions seen in cystic fibrosis – characterization of the cystic fibrosis transmembrane regulator provided a pathological mechanism which, in turn, may offer a therapeutic option.

Translational research is a two-way process which provides a bridge between fundamental research and applied clinical research and adheres to the philosophy of ‘bench-to-bedside and back again’. In the forward direction, it aims to generate supportive pre-clinical data prior to investigating the clinical application of:

• One of a number of different potential clinical biomarkers ( Box 37.3 )

  Box 37.3

  Clinical biomarkers

  • Prognostic biomarkers – used to stratify patients according to the prognosis of their disease subtype. In childhood acute lymphoblastic leukaemia, cytogenetic analysis identifies patients at a higher risk of treatment failure, e.g. t(9;22) Philadelphia chromosome.

  • Predictive biomarkers – predict a patient's response to a particular treatment. Mutations of Kir6.2 causing infant-onset diabetes insipidus predicts sensitivity to sulphonylureas ( Box 37.4 ).

    Box 37.4

    Interventional cohort studies provide an alternative to RCTs when they are unfeasible or unethical

    Scenario : You are a paediatrician working in a tertiary paediatric diabetes service.

    Framed PICO research question : In children with diabetes due to mutations within the ATP-sensitive potassium channel [population], is oral sulphonylurea [intervention] as effective as standard insulin replacement [control] in maintaining or reducing HbA1c [primary outcome measure] without additional episodes of hypoglycaemia [secondary outcome]?

    Method: This interventional cohort study looked at the management of 49 patients with diabetes due to mutations within the ATP-sensitive potassium channel.

    Results : A significant reduction in HbA1c from 8.1% pre-treatment to 6.4% after 12 weeks of treatment (p < 0.001). Forty-four patients were able to stop insulin treatment.

    Discussion: This collaborative multinational study demonstrated a safe and more effective way of managing diabetes resulting from this rare group of mutations. The patients in this study, all of whom were established on ‘standard’ insulin therapy, provided their own controls for the primary outcome measure of HbA1c, which can be assessed using a paired Student's t-test (see Chapter 38 , Statistics). Assuming that all patients were stable prior to enrolment, this approach provided the most appropriate trial design. It maximized the opportunity to study the effect of an intervention in a small population of patients with a rare condition. Randomizing patients between ongoing standard care and investigational treatment would have reduced the number in the trial therapy arm, thereby reducing the power of the study. However, a crossover approach would clearly not be appropriate in acute conditions where baseline measurements cannot provide stable data, or in life-threatening conditions where two interventions – usually current standard of care and new intervention – must be directly compared. In that situation, randomization would be most appropriate.

    Reference: Pearson ER, Flechtner I, Njølstad PR, et al. Switching from insulin to oral sulfonylureas in patients with diabetes due to Kir6.2 mutations. New England Journal of Medicine 2006:355;467–77.

  • Response biomarkers – provide a surrogate measure of a patient's disease status and response to the chosen therapy – fever or C-reactive protein in infection.

  • Pharmacokinetic biomarkers – used to assess the therapeutic or toxic effects of a drug. Antibiotics such as gentamicin or vancomycin will commonly have drug levels monitored.

  • Imaging biomarkers – non-invasive imaging, e.g. CT or MRI, may provide prognostic or response biomarker information.

• Novel therapeutic targets – if the CFTR gene is mutated in cystic fibrosis, is it possible to deliver a functionally normal gene to the airways and does this result in a biological improvement, e.g. in secretions?

In reverse, it aims to identify:

• Important clinical problems and frame them as research questions for fundamental investigation

• Unusual clinical situations as routes for fundamental science to develop learning of basic processes – identification of recessive genetic causes of primary immunodeficiency can identify critical elements in development of immunity.

Clinical research seeks to take potential diagnostic, monitoring or therapeutic strategies, identified by basic and translational research, into a representative clinical setting. It asks the question, ‘Does this treatment improve the health and well-being of real patients?’ Addressing this question will normally be performed in steps, or phases, with different aims at each phase ( Table 37.1 ). Phase I and II ( early phase ) trials recruiting small numbers of patients and focusing on determining the safety and appropriate dosing/schedule of a new intervention, may be combined so that patients are initially recruited to a phase I element followed by progression to a phase II element. This is increasingly common in trials of new therapies designed specifically for rare patient or disease/molecular subgroups.

Increasingly, experimental medicine (see Fig. 37.1 , and see Personalized medicine, below) seeks to use human trials to generate pre-clinical data. For example, a clinical trial with a primary aim of investigating the effect of a new cytotoxic anti-cancer drug can also be used to assess the pharmacokinetics/pharma­codynamics of that drug and the effect on similar, potentially cross-reactive pathways in normal tissue. These studies cannot be performed in vitro or in healthy human volunteers due to the potential toxicity.

Common trial designs

A number of different quantitative trial designs exist. Whilst the randomized controlled trial is frequently seen as the gold standard approach for investigating a new treatment intervention, each of the designs can be the most appropriate choice for a given clinical setting ( Table 37.2 ).

Qualitative research

Whilst most clinical research revolves around the quantitative analysis of clinical criteria, the interrogation of qualitative information can also play an important role, both in informing quantitative clinical trial design and in influencing clinical service design and provision.

Qualitative research ( Box 37.5 ) aims to develop understanding of a defined area by collecting information, analysing it and using the output to generate new ideas or hypotheses which may or may not then be suitable for quantitative analysis. Information collecting may involve:

• Review of documented evidence

• Observational approaches – recording uninfluenced behaviours

• Interviews – usually open-ended and defined by topics rather than specific questions

• Group discussion – specifically focusing on the interactions of the group setting

Whilst outcome measures are not restricted in qualitative studies, the study aim, methodology and analysis are no less rigorously defined and validated than in quantitative approaches. Commonly a number of validated methods will be used independently by more than one researcher and the results triangulated to identify areas of agreement, providing a high degree of validity.

Introduction – identifying a clinical need

The origins of all clinical trials should be an identifiable clinical need . This is one of the key contributions to be made by translational researchers. The process of identifying a need, investigating potential solutions and then validating them is key to framing the clinical research question ( Fig. 37.2 ). Whilst basic scientists may follow lines of investigation which appear interesting in the hope of further defining fundamental elements of biology, a greater expectation must be placed on the question underpinning a clinical trial. Asking a child and their family to consent to novel intervention or additional investigations can only be ethically justified if the benefit outweighs the risk. There must therefore be a reasonable expectation of patient benefit, identified as a clinical need. It may not, for example, be justified to randomize children between two established therapies with an equivalent, well-documented success rate, simply to demonstrate equivalence in a formal setting. If, however, there was a rationale for one treatment being more effective or less toxic than the other, then this would need to be investigated in a randomized clinical trial. The research question becomes:

• Treatment effect:

  • –

    Is treatment A more effective than treatment B in this clinical setting (e.g. In pre-school children with mild chronic asthma [patient], is oral leukotriene antagonist [intervention] as or more effective then inhaled steroid [control] in preventing chronic symptoms and/or acute exacerbations [outcomes])?

• Side effects/toxicity:

  • –

    Is treatment A less toxic than treatment B in this clinical setting (e.g. in pre-school children with mild chronic asthma [patient], does oral leukotriene antagonist [intervention] lead to greater height potential [outcome] than inhaled steroid therapy [control])?

Fundamental scientific evidence

Having identified a clinical need, basic evidence must be sought on how to approach meeting that need. Frequently this will involve a thorough search of the published literature, as there is often a large body of evidence in existence. However, this will often need to be complemented by additional studies investigating the relevance of published knowledge in the specific disease setting. Combined, these data will hopefully provide possible routes for a new intervention aimed at meeting the clinical need.

Pre-clinical modelling of the intervention

Having identified a potential novel intervention, this needs to be validated in the most appropriate pre-clinical model available. Some models will involve in vitro drug testing using derived cell lines. However, more complex models, often involving animal research, are frequently used as these may be felt to be more representative of the clinical setting with, for example, metabolism of drugs and complex tissue microenvironments in place of homogeneous single cell cultures ( Box 37.6 ). However, the animal model may only partially represent the human disease and the metabolism and cellular effects of a drug may differ between animal model and humans. One approach to solving this, predominantly relevant to malignancies, is to use human diseases engrafted into immunodeficient animal models – xenografts. It should be noted that even then, factors such as differential metabolism of a drug or the altered microenvironment may limit the clinical relevance of the model.

Defining the research hypothesis

Having identified a clinical need, defined the research question and examined the basic and pre-clinical evidence available, the approach to answering the question and resolving the need must be expressed as a hypothesis ( Box 37.7 ). A research hypothesis aims to state a prediction about the outcome of a research study, which can then be experimentally tested. Conventionally this hypothesis is based on the outcome of literature searches and previous, pre-clinical, studies. However, it is not statistically possible to prove the research hypothesis, as the results observed within the study might be due to chance occurrence. Instead, a study's results must be compared against the opposite situation, which is known as the null hypothesis . Evidence which argues against the null hypothesis is evidence in favour of the original research hypothesis, which now becomes known as the alternative hypothesis . Deciding what size of effect the study is looking to identify is an essential component of defining the alternative hypothesis, providing the clinically relevant outcome with which power calculations can be made and data analyses can be performed. Defining the research question and subsequently the research hypothesis is key to determining how the resultant study will be structured, performed and analysed.

Feasibility

Consideration must be given to the ability of the study team to deliver a successful trial within a suitable timeframe. The key factors to consider are shown in Box 37.8 . If the expected duration of the study, including necessary follow-up, risks making the delivery of results so delayed that they may no longer be clinically relevant (if, for example, other approaches to management have improved substantially), then it may not be appropriate to initiate the trial. Options for increasing recruitment should be considered, especially widening the trial setting to include national ( Box 37.9 ) or, commonly, international recruitment, the use of alternative trial designs and the sharing of data across trials by prospective or retrospective meta-analysis.

Box 37.9

Stepwise improvement in survival resulting from extensive recruitment to sequential childhood acute lymphoblastic leukaemia trials in the UK

Widespread recruitment to clinical trials within paediatric oncology has seen a stepwise improvement in outcome for many childhood malignancies. Improvements in supportive care have been an important part of this development, including:

• Criteria for identification of febrile neutropenia and the use of broad-spectrum antibiotics

• Improved prophylaxis for Pneumocystis jiroveci and population immunity for measles

• Improved imaging – CT, MRI, CT/PET, isotope imaging

• Improved surgical/anaesthetic techniques – operating microscopes, coagulation diathermy, peri/intraoperative imaging

• More advanced intensive care facilities

However, many of the clinical trials conducted in childhood malignancy have provided a clear improvement in outcome. This is well demonstrated by the results of UK trials in acute lymphoblastic leukaemia ( Fig. 37.3 ), which have recruited over 8500 children between 1980 and 2011, with the most recent complete trial, UKALL 2003, recruiting over 95% of eligible children:

• UKALL VIII – Attempted to reproduce the superior results seen in the US by using the US CCG162 protocol. Inclusion of daunorubicin in induction improved disease control, albeit at the cost of higher treatment related mortality. Extension of maintenance therapy from two to three years improved survival, although again at the cost of increased toxicity. Overall, similar results were achieved compared to the US, but neither randomization demonstrated a clear overall benefit.

• UKALL X – Examined the role of post-induction intensification. Inclusion of both early and late intensification phases was better than one or no intensification phase. Five year event-free survival was 71% with 2 intensification phases, 62–63% with a single intensification and 57% without intensification.

• UKALL 97/99 – This trial demonstrated the superiority of dexamethasone over prednisolone, stratified patients according to white cell count and age at diagnosis: standard risk – age <10 years, white cell count >50 × 10 ⁹ /L; high risk – age >10 years or white cell count >50 × 10 ⁹ /L. The UK trial again adopted an apparently superior US approach with longer intensification blocks. Introduction of more extensive and sensitive cytogenetic analysis to identify high risk groups.

• UKALL 2003 – Stratified children according to minimal residual disease analysis (MRD), a molecular test for low levels of residual disease not detectable by traditional bone marrow analysis. Low risk MRD allowed for reduction of treatment intensity whilst high risk MRD resulted in escalation of treatment.

Further reading :

• Hargrave DR, Hann II, Richards SM, et al. Medical Research Council Working Party for Childhood Leukaemia. Progressive reduction in treatment-related deaths in Medical Research Council childhood lymphoblastic leukaemia trials from 1980 to 1997 (UKALL VIII, X and XI). Br J Haematol 2001;112;293–9.

• Mitchell C, Richards S, Harrison CJ, Eden T. Long-term follow-up of the United Kingdom medical research council protocols for childhood acute lymphoblastic leukaemia, 1980–2001. Leukemia 2010:24;406–18.

Preliminary statistics

One critical component of developing a clinical trial is to give consideration to defining the outcomes to be assessed and the number of patients required to demonstrate whether those outcomes are affected by the intervention on trial. By defining the outcome which the intervention is hoped to produce, statistical analyses can be tailored to address that specific question. Collecting large quantities of data and then asking multiple questions is statistically unsound and will both risk identifying effects resulting from chance and reduce the likelihood of producing a statistically significant result for the most important clinical outcome as statistical significance needs to be adjusted from when performing multiple tests (Bonferroni calculation). Ideally, outcomes should be defined as:

• Primary outcome measures – the main outcome(s) under investigation

• Secondary outcome measures – additional important impacts of the intervention

Once the nature of the outcome to be assessed is defined, preliminary data on the clinically relevant size and distribution of the effect within the study population can be used to determine the number of patients (sample size) which need to be recruited to the study. This is the power calculation , which will usually be required by regulatory/funding bodies as one part of demonstrating that the study is practicable. The factors required to produce a power calculation are:

• The significance level to be used in the analysis – the value against which the likelihood of incorrectly rejecting the null hypothesis will be judged. This will be calculated from the trial data as the p-value .

• The magnitude of effect in the study population (either a direct measurement or, preferably, standardized according to the spread of the effect within the population)

• The power – the probability of rejecting the null hypothesis when the alternative hypothesis is correct ( Box 37.10 ). Frequently set to be 0.8, this represents an 80% chance of correctly identifying the pre-determined clinically significant effect in the study population.

  Box 37.10

  Example of a power calculation

  Scenario: You are a general paediatrician who regularly treats children with pneumonia (see Table 39.6 ).

  Framed PICO research question : In a pre-school aged child with pneumonia [patient], are oral antibiotics [intervention] as effective as intravenous antibiotics [comparison] for time to resolution of symptoms, rate of hospital admission, length of stay and rate of complications [outcomes]?

  Power calculation: You work with a steering group who advise a difference of more than 20% could not be considered clinically equivalent (magnitude of effect) .

  With a 5% level of significance (statistical significance) , 80% power and equivalence defined as no more than a 20% difference ( magnitude of effect ) between treatments of the proportion meeting the primary outcome measure at any time, 98 children would be required in each arm of the trial.

  Reference: Atkinson M, Lakhanpaul M, Smyth A, et al. Comparison of oral amoxicillin and intravenous benzyl penicillin for community acquired pneumonia in children (PIVOT trial): a multicentre pragmatic randomised controlled equivalence trial. Thorax 2007;62:1102–6.