Hypothesis 82

Primary hypothesis

Null Hypothesis: Chocolate rich in cocoa flavanols does not reduce

cardiovascular risk in individuals with T2DM and leads to weight gain as part of a clinical trial.

Chapter  Two:  General  Methods  and  Materials    

• To summarise experimental approaches used within this thesis.

• To consider the role and design of clinical trials in producing evidence for the efficacy of interventions in healthcare.

• To consider the role of the placebo/control in clinical trials of food products. • To consider the six types of marker that were defined in Chapter One for

assessing the effect of chocolate in type 2 diabetes mellitus (T2DM).

• To consider the safety assessments and participant reported outcome measures. To address these aims this chapter will be split into three sections:

2.1 - Introduction to study design.

2.2 - A review of clinical trial methodologies relating to food. 2.3 - General procedures and methodologies.

2.1  –  Introduction  to  Study  Design  

An exploratory review (Section 1.4), with limited meta-analysis (Chapter 3), on the effect of chocolate on measures of diabetes control and cardiovascular risk was undertaken first to determine gaps in the knowledge base. This meta-analysis focussed on biological markers of health in one of six areas noted in Chapter One; Section 1.3.3:

1. Blood pressure 2. Lipid profile

3. Diabetes control and insulin resistance 4. Endothelial function

5. Oxidative Stress 6. Inflammation

In addition to these measures, weight change was determined as a marker for the additional energy load consumed that may adversely affect diabetes control.

Following this meta-analysis de novo clinical trial data was obtained from three randomised clinical trials in participants with T2DM summarised in Table 2.1.1, namely:

1. A pilot proof of concept study (Chapter Four – Study One).

2. An acute response study using transient hyperglycaemia to induce metabolic stress (Chapter Five – Study Two).

3. A three-arm double-blinded randomised controlled trial, which included a flavanol, enriched milk chocolate (Chapter Six – Study Three).

Table 2.1.1: Summary of study protocols, which constitute the experimental data presented within this thesis.

Exploratory Review and Meta- analysis Pilot Proof of concept study Acute response study Three-arm study

Chapter Chapter Three Chapter Four Chapter Five Chapter Six

Design Exploratory Review

and Meta-analysis Randomised Controlled Trial Double blind Intervention Cross-over Randomised Controlled Trial Double blind Intervention Cross-over Randomised Controlled Trial Double blind Intervention Parallel Regulatory framework PRISMA Guidelines (Adapted for a single author due to the nature of this piece of

work, search limited and estimation of variance of mean difference limited) ICH NHS ethics/ NHS Research & Development Trials registration ISRCTN 25655161 ICH NHS ethics/ NHS Research & Development Trials registration ISRCTN 35988358 ICH NHS ethics/ NHS Research & Development Trials registration NCT01617603

Funding None Diabetes Research

funds / chocolate provided as a gift

Barry Callebaut Nestlé 07.52.NRC

Publication One Abstract only to date (Accepted) Not suitable in current

state for publication due to limitations in methodology Mellor et al., (2010) + 3 abstracts Mellor et al., (In press) + 2 abstracts

Three abstracts only to date

Participants RCTs with well- defined interventions T2DM (lifestyle or oral medications only) T2DM (lifestyle or metformin only) T2DM (lifestyle or metformin only)

Duration All 8 weeks (2 months)

+ 4 weeks (1 month) washout 3 hours + 1 week washout 12 weeks following a 4 week run in period

Chocolate Chocolate or Cocoa in any dose Control matched in terms of appearance/

taste and energy/ macronutrients 45g per day Formulated high polyphenol chocolate Dyed white chocolate 13.5g dose Acticoa (3% polyphenol) Acticoa (0.9% polyphenol) 20g per day (40g acute feeding dose)

Milk chocolate (20mg epicatechin) Nestlé Noir (20mg epicatechin) Chocolate (1mg epicatechin) Primary Outcomes Effects of chocolate or cocoa on health Improvement in lipid profile Improved endothelial function Reduced insulin resistance Secondary Outcomes Chocolate showing superior effect compared to cocoa Weight change Glycaemic profile Blood pressure Insulin resistance Oxidative stress, insulin and glycaemic response

Lipid profile, blood pressure, inflammation and

2.2-­‐  Clinical  Trial  Design  

2.2.1-­‐  Hierarchy  of  Clinical  Evidence  

Traditionally when evaluating and generating data regarding evidence for health outcomes, a hierarchy of evidence has been developed (Ho, Peterson & Masoudi, 2008). This is often represented as a pyramid (figure 2.2.1.1), which is capped with a further tier of systematic reviews and meta-analyses of well-designed randomised controlled trials. The evidence for how this shapes the current evidence can be seen in Chapters One (introduction and literature review) and Three (exploratory review and meta- analysis). For the purposes of this chapter, it is being used to highlight the strength of the methods selected for the generation of the new data, which are presented in Chapters Three, Four, Five and Six.

Figure 2.2.1.1: Pyramid or hierarchy of clinical evidence (adapted from Ho, Peterson and Masoudi 2008).

The selection of an exploratory review with meta-analysis for Chapter Three is to consider the weight of the data in the peer-reviewed literature, and how the experimental data presented in the preceding three chapters potentially influences it. This is partially to confirm previous meta-analyses findings of a benefit of cocoa and chocolate, although a number of these have included mixed interventions, including

other biologically active ingredients or exercise. The secondary outcome is perhaps more important; a sub-group analysis investigating whether chocolate offers greater efficacy compared to cocoa.

2.2.1.1  -­‐  Randomised  Controlled  Trials  

Although observational studies, supported by the mechanistic studies provided by basic scientific work, there is no clinical trial data to date, available which supports a beneficial causative effect of chocolate in individuals with T2DM.

2.2.2-­‐  Clinical  Trial  Conduct  

Conduct and reporting of clinical trials is well described in the case of those using pharmaceutical agents or investigational medicinal products. In Europe and the United States of America, these are tightly regulated by legislation. This is closely linked to the guidelines developed by the International Committee for Harmonisation (ICH), which set about the development of good practice, and ethics in medical research (ICH, 2005). However, when considering the work reported in this thesis, which investigated the health effects of foods or other nutritional products; no such regulations, guidelines or legislation exist.

The absence of regulation is in part circumvented by the recommendations of the International Committee of Medical Journal Editors (ICMJE) in 2005 who stipulated that all clinical trials prior to submission should have their protocol registered. This has become part of research culture in the United Kingdom, where National Health Service Research Ethics Committees (NRES) strongly consider it an ethical obligation, but not a statutory stipulation, that clinical trials should be registered (NRES, 2009). This also has its basis in pharmaceutical trials. In the USA, following the passing of the Food and Drug Administration (FDA) Amendment Act 2007 (FDA, 2007), all trial results needed to be registered.

The requirement is not simply aimed at pharmaceutical research targeting publication in highest quality medical journals as these recommendations have since been adopted by approximately a thousand titles. Additional what is considered to be a clinical trial has been expanded, since June 2007 the ICMJE adopted the WHO definition of clinical trial as:

“Any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions to evaluate the effects on health outcomes.” Where the meaning of a health-related intervention includes any intervention used to modify a biomedical or health-related outcome. This does not just include pharmaceutical agents, but extends to surgical procedures, devices, behavioural treatments, dietary interventions, and process-of-care changes. The health outcome being assessed includes any biomedical or health-related measures obtained in patients or participants, including pharmacokinetic measures and adverse events (ICMJE, 2009). As trial registration is generally seen as best practice, this was completed for all the clinical trials reported in this thesis, as shown in table 2.1.1. This is especially important for trials investigating interventions linked to potential intellectual property (not only pharmaceutical agents, but also medicinal devices and foods amongst others) where there is potential for commercial exploitation and therefore a tendency not to report negative findings. So, even if a trial is not fully reported in the form of a peer-reviewed paper, the registry entry will remain in the public domain.

Although not a legal requirement, studies were conducted in accordance with ICH Good Clinical Practice Guidelines (1996) as would be expected in accordance with European law (European Union, 2001; 2005) and then their subsequent adoption into UK law (Statutory Instrument, 2004; 2006a; 2006b) for trials involving investigational medicinal products. The rationale for this was that the centre in which the work was undertaken was primarily engaged in pharmaceutical trials, and the nature of these

guidelines meant the conduct and data quality would be of the highest possible standard. As two of the studies were commercially funded but investigator led, it meant that the study protocol (although independently designed by myself with the supervisory team) together with the data, subject was to a high degree of scientific scrutiny throughout the data collect and analysis. This included full trial monitoring and blinded data scrutiny in the case of both of these trials funded by Barry Callebaut BV and Nestlé.

2.2.3   -­‐  Parallel  Compared  with  Crossover  Design  

All studies followed a controlled experimental design, which assessed changes in anthropometric and biomedical markers along with the use of quantitative questionnaires. This method was chosen so to define the study population, also it is excepted that observational and cohort methodologies would not be sensitive enough to show the subtle changes that would be expected within the time frame of the thesis (six years). It is also considered that only randomised controlled trials have the ability to demonstrate a potential causative effect, with observational studies limited to suggesting associations (Ho, Peterson & Masoudi, 2008).

Cooper et al. (2008) suggested that future studies investigating the health effects of cocoa and chocolate should employ a crossover design as part of a randomised controlled study. Ideally this should also include a placebo or control, which, is blinded to both the participant and the investigator. The latter aspect will be considered later in this chapter (2.2.2.10).

The first two clinical trials (Chapters Four and Five) attempted to follow the recommendations of Cooper et al. (2008). These were chosen as the recommendations to follow, for a number of logistical reasons in addition to enhancing their scientific validity. The primary one being that it was a way of undertaking the study within the constraints of the initially limited available population. This changed through the course of the thesis following the initiation of a potential trial participant database developed in

conjunction with the local diabetes network. Also considering the effect size seen in the studies reviewed in Chapter One, and whether the effect may be of a lesser extent in individuals with T2DM, it was felt that using individuals as their own control would be likely to reduce the risk of inter-participant variation at baseline. This was felt to be most likely to have occurred in the study reported in Chapter Five, as this involved an oral glucose challenge, as glycaemia can vary between individuals. Therefore, it was felt that a crossover methodology was vital as it would be almost impossible to match individuals in terms of their glycaemic response prior to randomisation and that could lead to a large number of individuals being excluded, resulting in potentially lengthy and expensive screening procedures, due to inter-individual variation.

Despite the favouring of crossover-designed studies, they are not without their limitations and sources of confounding. Lathyris, Trikalinos and Ioannidis (2007) in a meta-analysis of crossover arm studies concluded both study designs tended to produce the same outcome, although the crossover design tended to show a more conservative estimate of effect. The parallel arm design tended to be limited when sample sizes were small. The crossover design is limited, in that although it may require fewer participants, they take longer to complete, generally a washout period is required between interventions and it assumes the participants are clinically and biologically stable throughout the study.

The issue with respect to the length of the washout period, is estimated from the half- life of the interventional product, which in the case of cocoa flavanols is short, with biological activity no longer evident after 22 hours (Spadafranca et al., 2010). However, biological markers being assessed may not revert to baseline for several weeks, e.g. glycosylated haemoglobin (HbA1c), which has a half-life of 28.7 days (Allgrove & Cockrill, 1988). Therefore for acute effect studies a minimum washout of 24-48 hours is required, and for chronic studies four weeks would be logical if including assessments

of glycaemic control. This would mean that the minimum length for a crossover study would be five months (20 weeks), which would allow two half-lives to see an effect upon HbA1c, and one half-life period for the washout. So when considering the inclusion criteria; if participants need to have been on stable medication for at least three months prior to enrolment, this would require participants to be clinically stable for a total of least eight months. With the nature of diabetes being a progressive disease, it may mean that some of the change seen might be the effect of the disease rather than lack of effect of the intervention or its potentially negative effect. An example of the progressive nature of diabetes, could be the increase in estimated cardiovascular risk after 1 year seen with the control intervention, as the intervention prevented this progression (Curtis et al., 2012).

The final consideration required in crossover-designed studies is to fully account for potential order effects. Theoretically, if the study is adequately randomised this should not be seen. Theoretically crossover studies can be confounded by carry over effects from one arm to the other. This can be controlled for statistically but with careful randomisation this should not be necessary.

In the case of food trials including the type presented in this thesis, which employed chocolate as the intervention product it could be considered that unlike drugs, which are often in a pill form and who’s form result in little or no psychological or sociological attachment by the participant, food might evoke an emotive response. Therefore in clinical trials using foods a crossover methodology this could introduce a degree of confounding due to the emotional attachment of individuals to one of the interventional foods. It is possible that participants might have preconceived ideas from the media about the chocolate they expect to consume and that this might reflect how they respond to both the proposed ‘active’ and ‘placebo’ bars.

The potential participant response to having two different chocolates in a crossover study was one of the reasons why in the third study (Chapter Six) a parallel-randomised design was used. This avoids bias from preconditioning which is potentially seen in a crossover study design, although biases from previous life experiences could not be eliminated.

Due to the nature of the funding of the studies presented within this thesis, the chocolate formulations used were different in each study (summary characteristics of the chocolates are shown in Table 2.2.7). All three studies were doubly blinded at the point of randomisation, although there were subtle differences between the suggested ‘active’ chocolate and the placebo or comparator chocolate which may lead to the participants becoming unblinded. The investigators encouraged participants not to discuss their thoughts about the study chocolate until the end of the study in order to reduce biasing influences towards the investigator. The potential bias of participants ‘unblinding’ themselves was partially addressed using taste trials in the studies reported in Chapters Four and Five. For the parallel designed study in Chapter Six, Nestlé as a sponsor did not see this as an issue, so no taste trials were conducted. The rationale being that a milk chocolate was one of the ‘active’ interventions so would not be as easily detected as a traditional high polyphenol dark chocolate, which has a bitter taste. The risk was also moderated by the studies parallel design.

2.2.3.1  -­‐  Randomisation  

All three studies were randomised prior to the enrolment of the first participants. The provider of the chocolate held the randomisation codes in all three studies. In the cases of Chapter Four (Study One) and Chapter Six (Study Three) this was Nestlé, Nestec York, UK and Nestlé Research Centre, Lausanne, Switzerland respectively and for Chapter Five (Study Two) Barry Callebaut BV, Lebbeke-Weize, Belgium. The

concealment of the randomisation was maintained until the initial analysis of the data was complete, at which point the code was broken.

2.2.4  -­‐  Regulatory  Approval  

All studies underwent full ethical review by the Hull and East Riding NHS Local Research Ethics Committee, and obtained approval prior to their commencement. Research governance approval was obtained through the Research and Development Department of Hull and East Yorkshire Hospitals NHS Trust.

All studies were conducted in accordance with the Declaration of Helsinki of the World Medical Association (WMA, 2000) and Good Clinical Practice (EMEA, 2002). The work in Chapter Four was supported by an unrestricted gift of the chocolate for the study by Nestec PTC, York, UK and was funded through the Diabetes Charitable fund. Funding and sponsorship from Barry Callebaut BV, Lebbeke-Weize, Belgium, supported the work in Chapter Five, although the author (along with the supervisory team) of the thesis was responsible for the trial design. The work in Chapter Six was supported, funded, sponsored and monitored by Nestlé Research Centre, Lausanne, Switzerland. The protocol and study design for the work presented in Chapter Six was developed by the author of the thesis in conjunction with their supervising team with support from the Nestlé Research Centre, Lausanne, Switzerland.

All three trials were registered on public data bases of randomised clinical trials; the work in Chapter Four being ISRCTN 25655161, Chapter Five; ISRCTN 35988358 (both available from http://www.controlled-trials.com/isrctn/) and Chapter Six; NCT01617603 (http://clinicaltrials.gov/ct2/home).

2.2.5  -­‐  Population  

Participants in all three studies had T2DM and were taking (if applicable) stable medication for at least three months prior to enrolment with different patients invited to

participate in each studies. Further details and of the inclusion and exclusion criteria are described in Chapters Four, Five and Six. All participants were recruited from within the Hull and East Riding Diabetes Network in which they were patients.

Participants had all given consent to be contacted regarding research when they were referred to the networks diabetes programme. It is a recommendation that all individuals with diabetes receive structured education about how to self-manage their condition (NICE, 2003, 2008; Diabetes UK & Dept. of Health, 2005). This was selected as a key criterion, as all participants would have received the same basic information regarding their diabetes and the recommended dietary treatment of their diabetes, thus reducing lifestyle variability issues.

2.2.6  -­‐  Demographics  

The population age range for all three studies was between 40 and 80 years. To reduce the potential for confounding from insulin deficiency, apart from the initial study (Chapter Four), which had a body mass index (BMI) range of 19-45 kgm-2, the subsequent studies focused on overweight and obese individuals with a BMI range of 25-40kgm-2_{. The limits on weight were largely set to avoid extremes of metabolism,}

including the sympathetic over activity and inflammation associated with morbid obesity (Pontiroli, Pizzocri, Paroni & Folli, 2006).

All studies included both male and postmenopausal female participants. The rationale for only including postmenopausal women was to reduce potential causes of variation in insulin resistance associated with cyclical changes in female sex hormones and currently relatively few premenopausal women have T2DM (Geer & Shen, 2009). For the work in Chapters Four and Five there was no allocation of the proportion of males to females in the study. This was not considered to be a potentially significant