This article has Open Peer Review reports available.
Protocol for SAMS (Support and Advice for Medication Study): A randomised controlled trial of an intervention to support patients with type 2 diabetes with adherence to medication
© Farmer et al; licensee BioMed Central Ltd. 2008
Received: 03 March 2008
Accepted: 11 April 2008
Published: 11 April 2008
Although some interventions have been shown to improve adherence to medication for diabetes, results are not consistent. We have developed a theory-based intervention which we will evaluate in a well characterised population to test efficacy and guide future intervention development and trial design.
Methods and Design
The SAMS (Supported Adherence to Medication Study) trial is a primary care based multi-centre randomised controlled trial among 200 patients with type 2 diabetes and an HbA1c of 7.5% or above. It is designed to evaluate the efficacy of a two-component motivational intervention based on the Theory of Planned Behaviour and volitional action planning to support medication adherence compared with standard care. The intervention is delivered by practice nurses. Nurses were trained using a workshop approach with role play and supervised using assessment of tape-recorded consultations. The trial has a two parallel groups design with an unbalanced three-to-two individual randomisation eight weeks after recruitment with twelve week follow-up. The primary outcome is medication adherence measured using an electronic medication monitor over 12 weeks and expressed as the difference between intervention and control in mean percentage of days on which the correct number of medication doses is taken. Subgroup analyses will explore impact of number of medications taken, age, HbA1c, and self-reported adherence at baseline on outcomes. The study also measures the effect of dispensing medication to trial participants packaged in the electronic medication-monitoring device compared with conventional medication packaging. This will be achieved through one-to-one randomisation at recruitment to these conditions with assessment of the difference between groups in self-report of medication adherence and change in mean HbA1c from baseline to eight weeks. Anonymised demographic data are collected on non-respondents. Central randomisation is carried out independently of trial co-ordination and practices using minimisation to adjust for selected confounders.
The SAMS intervention and trial design address weaknesses of previous research by recruitment from a well-characterised population, definition of a feasible theory based intervention to support medication taking and careful measurement to estimate and interpret efficacy. The results will inform practice and the design of a cost-effectiveness trial [ISRCTN30522359].
Diabetes is a major public health problem. The number of people with diabetes is estimated to reach 330 million by 2030 . There is a high clinical and economic burden from the disease : people with diabetes have a two-to-four fold increased risk of cardiovascular disease compared to the general population and increased incidence of retinopathy, peripheral nerve damage and renal problems.
Evidence supports the use of multiple medications to control blood glucose and cardiovascular risk among patients with type 2 diabetes , and, this may lead to prescriptions of eight or more medications a day. Up to half of this medication may not be taken as prescribed [4, 5], with adherence to medication falling as dosage frequency rises . Failure to take medication has important consequences. It not only reduces efficacy of the treatment, but wastes healthcare resources in prescribed pills not taken, extra consultations, referrals, investigations and hospital admissions [7, 8]. The availability of an effective intervention to support patients with type 2 diabetes in taking their medication regularly would make a major contribution to human health.
A variety of interventions to support adherence to medication have been tested for efficacy , and their clinical application assessed . Although some interventions are effective in improving adherence, the results are inconsistent. These studies have used complex packages of care, targeting multiple self-care activities and therefore limiting identification of factors that lead to and might modify non-adherence. Approaches to supporting behaviour change are now informed by evidence about the psychological determinants of behaviour, and techniques to alter them, and there is potential to apply these techniques in the field of medication adherence [11, 12].
Adherence to medication can be defined in many ways. Our definition is the extent to which medicines are taken regularly as prescribed. We argue that to understand the determinants of adherence better and optimise the impact of different intervention components upon them, we need to design interventions based on clear conceptual frameworks. We also need to test whether targeting determinants of medication adherence (e.g., patient's beliefs) results in greater levels of adherence [11, 13].
People may not take their medication because of ambivalence about pros and cons of adherence (intentional lapse). Others, who intend to take their medications regularly, may still forget to do so (non-intentional lapse). Intentional and non-intentional lapses in adherence have been identified as two important elements contributing to overall non-adherence . Drawing on psychological theory and evidence [15, 16], we have developed approaches to address these two elements: to increase patients' motivation to take their tablets regularly by targeting underlying beliefs; and to help patients define specific action plans to facilitate the translation of intentions into action and habit formation. The two components can be defined as motivational – targeting the cognitive determinants of intention – and volitional – defining action plans to implement the target behaviour .
The motivational component of the intervention is based on the Theory of Planned Behaviour (TPB) . This theory specifies beliefs which may be targeted to strengthen intention. In particular techniques include reinforcement of positive beliefs and problem solving approaches to negative beliefs. The theory's constructs account for 35 to 50% of variance in intention and 26 to 35% of variance in behaviour . This illustrates a gap between intention and behaviour. The use of action planning, also called implementation intentions, is a promising approach to bridge this gap. A range of studies has demonstrated that explicit consideration of the circumstances under which a behaviour will be enacted (action plans), promotes clinically important change in specific health related behaviours including consumption of vitamin C tablets, attendance for cervical cytology screening and breast self-examination .
In clinical trials evaluating medication taking, adherence in the control group is often high. This may be because patients willing to participate in trials are likely to be adherent, because self-report measures may overestimate true adherence, or because more objective measurement focuses attention, on this target behaviour [7, 10]. To interpret the effects of interventions better, future trials need to characterise the population from which participants have been recruited by collecting data about non-responders, encouraging respondents who do not wish to take part in the trial to complete baseline questionnaires, and investigating the effect of measurement itself on adherence and clinical risk.
In previous pilot work we have developed measures and interventions  using the approach defined by the Medical Research Council Framework for development of interventions for evaluation in randomised trials [20, 21]. We therefore intend to estimate the efficacy of support for medication adherence on medication taking behaviour using motivational and action planning techniques; and estimate the effect of monitoring medication-taking itself on self-reported adherence and glycaemic control in a randomised trial. The trial addresses important limitations in the literature by recruitment from a well-characterised population, definition of a feasible intervention to support medication taking, and careful measurement to estimate and interpret efficacy.
The London, multi-centre research ethics committee has reviewed and approved the protocol (06/MRE02/3).
Recruitment of patients is from general practices in Oxfordshire, Milton Keynes, Suffolk, Essex and Huntingdonshire.
Patients are eligible for inclusion in the trial if diagnosed aged 18 years or above with type 2 diabetes of at least three months duration, able to give informed consent, currently taking any oral glucose-lowering agent and with a HbA1c ≥ 7.5%. Those approached are deemed by their general practitioner able to complete a consent form, independent in medication taking, and appropriate for tight glycaemic control.
After an eight week period of initial observation, patients are allocated to either an intervention visit to support medication adherence, or to a control visit at which their current medication regimen is recorded.
The intervention was developed and piloted after a detailed study to identify beliefs held about diabetes . The intervention visit is delivered by the practice nurse. In the first, motivational component of the intervention, the nurse elicits the patient's beliefs relevant to their intention to take medication regularly as prescribed based on the TPB. Nurses use scripts to guide the order and phrasing of questions that elicit individual beliefs about benefits and harms, views of important others and factors that may facilitate or inhibit taking medication regularly as prescribed. Positive beliefs are reinforced with provision of tailored information and problem solving is applied in relation to negative beliefs . In the second, action planning component, the nurse asks the patient to write down the exact circumstances under which they will take their medication (using an "if.then" formulation to elicit where, when and how this will occur) . In the control visit, delivered by the practice nurses, none of the above techniques are applied.
Initial training is provided over one day for the practice nurses delivering the intervention, using a workshop approach with role-play led by a psychologist and intervention facilitators involved in quality assurance of intervention delivery. It includes explanation of the rationale for the intervention, use of a detailed manual providing information about the theoretical base and operationalisation of the different components, how to use the intervention to establish agreement with the prescription plan, and use of a 'script' to promote fidelity of intervention delivery. Assessment of audiotapes of intervention sessions is conducted to assess fidelity of delivery using standardised forms. Subsequently, coaching and feedback is provided to the nurses to optimise intervention delivery. Possible sources of bias in intervention delivery are addressed with the practice nurses throughout training and ongoing support. These include the need to avoid (i) delivering the intervention to control group patients, (ii) discussing the intervention techniques with other members of the primary care team and (iii) using intervention techniques not specified in the script.
At the baseline visit patients are randomised to be dispensed their metformin tablets in a container with a lid that records the occurrence and timing of opening (TrackCap, Aardex, Zurich, Switzerland), or to continue taking their medication in standard packaging as before the trial. If participants are not taking metformin, another of their usual oral hypoglycaemic medication is dispensed in the electronic medication-monitoring device.
Randomisation of patients to assess intervention and baseline measurement effects is carried out independently of trial co-ordination and intervention by the trial statistician. A partial minimisation procedure is used to dynamically adjust randomisation probabilities to balance the baseline stratification variables. For measurement effect, randomisation at baseline includes practice, duration of diabetes, HbA1c result from the practice record and patient baseline measure of self-report adherence. For intervention, randomisation at eight weeks includes the baseline measurement-effect randomisation group allocation, and baseline HbA1c is included instead of the practice record. Staff receiving and downloading the data from the electronic medication-monitoring device are blind to allocation, as are laboratory staff measuring HbA1c.
Baseline data collection and randomisation to electronic medication taking measurement
Trial measures and timing
Before 8 week visit
At 8 week visit
After 8 week visit
Final visit at 20-weeks
HbA1c (Central laboratory)
Electronic medication monitor for preceding period
Self-report (MARS) Diabetes
TPB† Direct measures and intention
TPB† Indirect measures
Time and location of taking medicines
Communication with practice nurse
Quality of Life
Diabetes Treatment Satisfaction Questionnaire
Theory of Planned Behaviour measures of belief and intention towards taking oral glucose-lowering medications
Number of items†
It is beneficial for me to take my diabetes medicines without missing a day
Direct subjective norm (injunctive)
Most people who are important to me think I should take my diabetes medicines without missing a day
Direct subjective norm (descriptive)
If they were taking part in this study, most people who are important to me would take their diabetes medicines without missing a day
Direct perceived behavioural control
It is difficult for me to take my diabetes medicines without missing a day
If I were to take my diabetes medicines without missing a day, it would keep my diabetes under control
Indirect subjective norm (injunctive)
Members of my family or close relatives would approve of me taking my diabetes medicines without missing a day
Indirect perceived behavioural control
Changes to my daily routine would make it more difficult for me to take my diabetes medicines without missing a day
I intend to take my diabetes medicines without missing a day
Willing patients attend a baseline recruitment consultation of 40 minutes with their practice nurse. In advance of the visit patients are allocated to be dispensed medication in the electronic medication-monitoring device or in standard packaging. Informed consent is obtained and additional clinical data collected. Blood tests are taken, and questionnaires completed. For those allocated to the electronic medication-monitoring device, its use is explained, and the practice dispenser or pharmacist dispenses the patient's usual prescription for metformin or alternative oral glucose lowering agent in the device. For those allocated to standard packaging, the practice dispenser or pharmacist dispenses medication in standard blister-packs. A follow-up visit is arranged in eight weeks and, in advance of the visit, the patient is randomised to be allocated to receive the intervention or control visit.
Intervention and control visit
Prior to the eight-week study visit patients are sent a questionnaire to be completed at home and returned to the coordinating centre in a sealed envelope by post. At the eight-week visit, patients allocated to the intervention take part in a 50 minutes interview with the practice nurse including the intervention and data collection. The control visit lasts approximately 20 minutes in which the data are collected. All patients at the eight week visit have blood samples taken and inquiry made about any possible adverse events including hypoglycaemia. A postal questionnaire is completed one week after the eight week visit.
Follow up to 20 weeks
Final follow up for all patients at 20 weeks involves a visit to the practice nurse and includes a blood sample for measurement of HbA1c and drug concentrations, and a final questionnaire.
Primary outcome for intervention trial
The primary outcome for the intervention is the percentage of days on which the prescribed dose of main hypoglycaemic medication is taken as prescribed, measured with the TrackCap electronic medication monitor over a 12-week period from eight to twenty weeks.
Primary outcome for measurement-effect trial
The primary behavioural outcome for the measurement effect of the use of an electronic medication-monitoring device compared with standard packaging is change in self reported adherence measured with the Medication Adherence Report Schedule (MARS). The primary clinical outcome for the measurement effect is mean HbA1c at eight weeks.
Secondary outcomes for intervention trial
Secondary outcomes are HbA1c, well-being measured with the 12-item Short Form Medical Outcomes Study health survey questionnaire (Short Form-12)  and treatment satisfaction measured with the Diabetes Treatment Satisfaction Questionnaire (DTSQ) .
Additional measures of medication adherence
Other measures include serum drug concentrations and self-reported measures of medication adherence (MARS) , and dispensing records to complement electronic medication-monitoring. Drug concentrations will be collected to explore whether the electronic medication-monitoring device records of timing of doses relate to the presence of therapeutic concentrations at study visits.
Psychological and process measures
A questionnaire based on the Theory of Planned Behaviour is used to assess attitude, subjective norm, perceived behavioural control, and intention to take diabetes medicines without missing a day. The questionnaire also assesses perceived consequences of taking diabetes medication (behavioural beliefs); perceived views of significant others about taking diabetes medication (normative beliefs); perceived factors that make taking diabetes medication easier or difficult (control beliefs). These measures are based on our earlier pilot work .
We are using a measure to assess patients' ratings of communication with nurses (covering the ability to tell the nurse personal or troubling things and feeling understood) , and ask open-ended questions about when and where medications are usually taken. Tape recordings of consultations are used to assess fidelity of intervention delivery, using a-priori criteria. The intervention facilitators record adherence to techniques specified in the intervention script on standardised forms.
Baseline measures to characterise participants
Socio-demographic and clinical measures include measures of duration of diabetes, overall drug regimen and numbers of prescribed medication doses. For the purposes of characterising the eligible population and randomisation at baseline, the last practice-recorded HbA1c is also recorded.
The trial is planned to follow-up 200 patients, providing 80% power at the 5% significance level to detect a difference in means between randomised groups of 5% (1.5 days per month difference) in the percentage of days on which the correct number of doses is taken. This was based on an estimate of the standard deviation of this measure of 13.5% in a pilot study for the trial conducted in 2001 in Newmarket, Cambridgeshire (personal communication Dr Simon Griffin). Assuming a follow-up rate of 85%, a maximum of 250 participants would need to be recruited.
For the measurement effect we can detect a difference of one point in the MARS self-reported adherence measure (a small to moderate effect size) , with 80% power and a two sided test at the 5% level with a conservative estimate of a 2.5% standard deviation. We will also detect a 0.5% difference in HbA1c at eight weeks with 80% power at the 5% level assuming a standard deviation of 1.25% for HbA1c.
Analysis will be by intention-to-treat using multiple imputation for those with missing data by the method of Rubin . This will be supported by a sensitivity analysis reporting all cases where the primary outcome is available and using optimistic and pessimistic scenarios with the imputed estimates for missing data . The mean percentage of days on which the correct number of doses is taken will be compared between the groups allocated to intervention and control, using a non-parametric bootstrap method to derive the difference in means with a 95% confidence interval [28, 29].
The primary behavioural outcome for the assessment of measurement effect is self-reported adherence (MARS) at eight weeks compared between groups allocated to use of the medication monitor compared to those allocated to dispensing of medicines in standard packaging. The primary clinical outcome is HbA1c at 8-weeks compared between groups. In order to increase the precision of estimated intervention effects continuous measures will be analysed with adjustment for the baseline covariate prior to the first randomisation, if available, with missing baseline data included by the missing indicator method .
Subgroup analyses will be carried out to explore the impact of number of medications, age, HbA1c, self-reported adherence at baseline and prior randomisation to the electronic medication-monitoring device on the intervention effect.
We will assess the impact of the intervention on beliefs about taking diabetes medication, and test whether any observed change in medication adherence is mediated by the intervention's impact on beliefs. We will explore the potential role of nurse communication style as a moderator of effect. We will compare the differences between participating and non-participating patients.
This trial is designed to estimate the efficacy of a two-component intervention targeting determinants of intentions and action planning as possible mediators of improved medication adherence. Measures and interventions have been developed using the approach defined by the Medical Research Council Framework for development of interventions for evaluation in randomised trials [20, 21]. In addition to estimating efficacy the trial will provide information on psychological mechanisms of increasing medication adherence and assess the impact of trial measurement on outcomes.
We are conducting this trial in a well characterised primary care population. The training of nurses to deliver the intervention uses scripts and feedback from taped intervention sessions to maximise delivery as planned and consistent delivery across nurses and over time. We are able to test impact of the communication style of the practice nurse by examining the relationship between patient attitudes about communication and behavioural outcomes. We have optimised clear communication of the intervention through training and feedback to the nurses delivering the intervention.
This trial has been designed to address the weaknesses of previous work in the area of medication adherence. It approaches medication taking as a discrete behaviour and uses psychological theory and evidence in the design of intervention to address intentions and the step from intention to actions, and in measurement of the process. It will therefore be possible to explore the extent to which interventions that address these components of medication adherence might improve outcomes. Additional analyses will estimate the extent to which any observed effects of the intervention on behaviour are mediated by intention and action planning. The trial will explore the extent to which previous trials may have failed to show efficacy as a result of the kind of people who participate in such trials, and the kinds of measures of adherence used, including the impact of measurement itself on adherence and HbA1c.
This trial will provide one of the first evaluations of an intervention developed using psychological theory to support patients in adherence to diabetes medication. If the trial provides evidence of efficacy, it will add to the clinical approaches for medicines education currently in use.
If the two-component intervention shows evidence of efficacy, we propose to compare it with the action-planning only intervention in a future trial. Future research in this area will also need to incorporate detailed health economic assessments to assess cost-effectiveness. However, because of the large numbers of people with diabetes, even very small effects of an intervention on adherence offer potential for major public health gains.
This trial is supported by the Medical Research Council and through National Health Service R&D support funding.
The SAMS Team
Writing group: A Farmer, AT Prevost, W Hardeman, A Craven, S Sutton, S Griffin, A-L Kinmonth. Coordinating Centre: A Craven, J Oke, D White. Trial Statistician: AT Prevost. Intervention Development: A Farmer, W Hardeman, I Kellar, Y Kim, M Selwood, S Sutton. Measures: A Farmer, S Griffin, D Hughes, I Kellar, S Sutton. Practices: Suffolk; The Rookery Medical Centre Newmarket, Woolpit Health Centre; Huntingdonshire; Rainbow Surgery Ramsey, Ramsey Health Centre, The Surgery Papworth Everard, Spinney Surgery St. Ives, Eaton Socon Health Centre; Essex; John Tasker House Surgery Great Dunmow; Oxfordshire; Woodcote Surgery, Horse Fair Surgery Banbury, Islip Medical Practice; Milton Keynes; Parkside Medical Centre, Stantonbury Health Centre. Pharmacies: Lloyds Pharmacy Ramsey, Cox & Robinson Banbury; Yogi Pharmacy Great Dunmow; P & I Smith Bletchley, Cox & Robinson Bletchley, Lloyds Pharmacy Eaton Socon, McLaren Pharmacy New Bradwell. Nurse training and intervention co-ordination: S Boase, Y Kim. Fidelity assessment: J Argles, S Boase, P Gash, Y Kim, M Selwood. Primary care network liaison: J Graffy. Pharmaceutical adviser: S. Ashwell.
We are grateful to the patients; practice nurses; general practitioners and pharmacists for taking part in this trial. We are grateful to Professor R Horne for permission to use the MARS scale and Professor C Bradley for permission to use the DTSQ. The Departments of General Practice at the Universities of Oxford and Cambridge are partners in the NIHR National School of Primary Care Research. The SAMs study is a component of the MRC co-operative on the development and evaluation of innovative strategies for the prevention of chronic disease in primary care. The opinions expressed in this paper are not necessarily those of the Department of Health.
- Wild S, Roglic G, Green A, Sicree R, King H: Global Prevalence of Diabetes: Estimates for the year 2000 and projections for 2030. Diabetes Care. 2004, 27: 1047-1053.View ArticlePubMedGoogle Scholar
- Yach D, Stuckler D, Brownell KD: Epidemiologic and economic consequences of the global epidemics of obesity and diabetes. Nat Med. 2006, 12: 62-66.View ArticlePubMedGoogle Scholar
- Gaede P, Vedel P, Larsen N, Jensen GVH, Parving HH, Pedersen O: Multifactorial Intervention and Cardiovascular Disease in Patients with Type 2 Diabetes. The New England Journal of Medicine. 2003, 348: 383-View ArticlePubMedGoogle Scholar
- World Health Organisation: Adherence to long term therapies: evidence for action. 2003, Geneva, World Health OrganisationGoogle Scholar
- National Council on Patient Information and Education: Enhancing prescription medicine adherence: a national action plan. 2007, Bethesda, MD, National Council on Patient Information and EducationGoogle Scholar
- Claxton AJ, Cramer J, Pierce C: A systematic review of the associations between dose regimens and medication compliance. Clinical Therapeutics. 2001, 23: 1296-1310.View ArticlePubMedGoogle Scholar
- Winkler A, Teuscher AU, Mueller B, Diem P: Monitoring adherence to prescribed medication in type 2 diabetic patients treated with sulfonylureas . Swiss Med Wkly. 2002Google Scholar
- McDonald HP, Garg AX, Haynes RB: Interventions to Enhance Patient Adherence to Medication Prescriptions: Scientific Review. JAMA. 2002, 288: 2868-2679.View ArticlePubMedGoogle Scholar
- Kripalani S, Yao X, Haynes RB: Interventions to Enhance Medication Adherence in Chronic Medical Conditions: A Systematic Review. Arch Intern Med. 2007, 167: 540-549.View ArticlePubMedGoogle Scholar
- Osterberg L, Blaschke T: Adherence to Medication. The New England Journal of Medicine. 2005, 353: 487-497.View ArticlePubMedGoogle Scholar
- Hardeman W, Sutton S, Griffin S, Johnston M, White AJ, Wareham NJ, Kinmonth AL: A causal modelling approach to the development of theory-based behaviour change programmes for trial evaluation. Health Educ Res. 2005, 20: 676-687.View ArticlePubMedGoogle Scholar
- Weinman J, Petrie KJ, Moss-Morris R, Horne R: The illness perception questionnaire: a new method for assessing the cognitive representation of illness. Psychol Health. 1996, 11: 431-445.View ArticleGoogle Scholar
- Leventhal H, Weinman J, Leventhal EA, Phillips LA: Health Psychology: The Search for Pathways between Behavior and Health. Annu Rev Psychol. 2008, 59: 477-505.View ArticlePubMedGoogle Scholar
- Barber N: Should we consider non-compliance a medical error?. Qual Saf Health Care. 2002, 11: 81-84.View ArticlePubMedPubMed CentralGoogle Scholar
- Sheeran P, Orbell S: Implementation intentions and repeated behaviour: augmenting the predictive validity of the theory of planned behaviour. Eur J Soc Psychol. 1999, 29: 349-369.View ArticleGoogle Scholar
- Ajzen I: The theory of planned behaviour. Organizational Behavior and Human Decision Processes. 1991, 50: 179-211.View ArticleGoogle Scholar
- Sutton S: Determinants of health-related behaviours: Theoretical and methodological issues. The Sage handbook of health psychology. Edited by: S S, Baum A and Johnston M. 2004, London, SageGoogle Scholar
- Sheeran P: Intention-behaviour relations: a conceptual and empirical review. Eur J Soc Psychol. 2002, 12: 1-36.View ArticleGoogle Scholar
- Farmer A, Kinmonth AL, Sutton S: Measuring beliefs about taking hypoglycaemic medication among people with Type 2 diabetes. Diabet Med. 2006, 23: 265-270.View ArticlePubMedGoogle Scholar
- Campbell M, Fitzpatrick R, Haines A, Kinmonth AL, Sandercock P, Spiegelhalter D, Tyrer P: Framework for design and evaluation of complex interventions to improve health. BMJ. 2000, 321: 694-696.View ArticlePubMedPubMed CentralGoogle Scholar
- Campbell NC, Murray E, Darbyshire J, Emery J, Farmer A, Griffiths F, Guthrie B, Lester H, Wilson P, Kinmonth AL: Designing and evaluating complex interventions to improve health care. BMJ. 2007, 334: 455-459.View ArticlePubMedPubMed CentralGoogle Scholar
- Ware J, Kosinski M, Keller SD: A 12-Item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Med Care. 1996, 34: 220-233.View ArticlePubMedGoogle Scholar
- Lewis KS, Jennings AM, Ward JD, Bradley C: Health belief scales developed specifically for people with tablet-treated type 2 diabetes. Diabet Med. 1990, 7: 148-155.View ArticlePubMedGoogle Scholar
- Horne R, Weinman J: Predicting treatment adherence: an overview of theoretical models. Adherence to treatment in medical conditions. Edited by: Myers L and Midence K. 1998, London, Harwood Academic, 25-50.Google Scholar
- Kinmonth AL, Woodcock A, Griffin S, Spiegal N, Campbell MJ: Randomised controlled trial of patient centred care of diabetes in general practice: impact on current wellbeing and future disease risk. The Diabetes Care From Diagnosis Research Team. BMJ. 1998, 317: 1202-1208.View ArticlePubMedPubMed CentralGoogle Scholar
- Schafer JL: Multiple imputation: a primer. Stat Methods Med Res. 1999, 8: 3-15.View ArticlePubMedGoogle Scholar
- White IR, Thompson SG: Adjusting for partially missing baseline measurements in randomized trials. Stat Med. 2005, 24: 993-1007.View ArticlePubMedGoogle Scholar
- Vern TF, Brian DMT, Patrick R: The Impact of Dichotomization on the Efficiency of Testing for an Interaction Effect in Exponential Family Models. Journal of the American Statistical Association. 2004, 99: 822-831.View ArticleGoogle Scholar
- Carpenter J, Bithell J: Bootstrap confidence intervals: when, which, what? A practical guide for medical statisticians. Statistics in Medicine. 2000, 19: 1141-1164.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2296/9/20/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.