A systematic review of the clinical effectiveness and cost-effectiveness of sensory, psychological and behavioural interventions for managing agitation in older adults with dementia

BACKGROUND

Agitation is common, persistent and distressing in dementia and is linked with care breakdown. Psychotropic medication is often ineffective or harmful, but the evidence regarding non-pharmacological interventions is unclear.

OBJECTIVES

We systematically reviewed and synthesised the evidence for clinical effectiveness and cost-effectiveness of non-pharmacological interventions for reducing agitation in dementia, considering dementia severity, the setting, the person with whom the intervention is implemented, whether the effects are immediate or longer term, and cost-effectiveness.

DATA SOURCES

We searched twice using relevant search terms (9 August 2011 and 12 June 2012) in Web of Knowledge (incorporating MEDLINE); EMBASE; British Nursing Index; the Health Technology Assessment programme database; PsycINFO; NHS Evidence; System for Information on Grey Literature; The Stationery Office Official Documents website; The Stationery National Technical Information Service; Cumulative Index to Nursing and Allied Health Literature; and The Cochrane Library. We also searched Cochrane reviews of interventions for behaviour in dementia, included papers' references, and contacted authors about 'missed' studies. We included quantitative studies, evaluating non-pharmacological interventions for agitation in dementia, in all settings.

REVIEW METHOD

We rated quality, prioritising higher-quality studies. We separated results by intervention type and agitation level. As we were unable to meta-analyse results except for light therapy, we present a qualitative evidence synthesis. In addition, we calculated standardised effect sizes (SESs) with available data, to compare heterogeneous interventions. In the health economic analysis, we reviewed economic studies, calculated the cost of effective interventions from the effectiveness review, calculated the incremental cost per unit improvement in agitation, used data from a cohort study to evaluate the relationship between health and social care costs and health-related quality of life (DEMQOL-Proxy-U scores) and developed a new cost-effectiveness model.

RESULTS

We included 160 out of 1916 papers screened. Supervised person-centred care, communication skills (SES = -1.8 to -0.3) or modified dementia care mapping (DCM) with implementing plans (SES = -1.4 to -0.6) were all efficacious at reducing clinically significant agitation in care home residents, both immediately and up to 6 months afterwards. In care home residents, during interventions but not at follow-up, activities (SES = -0.8 to -0.6) and music therapy (SES = -0.8 to -0.5) by protocol reduced mean levels of agitation; sensory intervention (SES = -1.3 to -0.6) reduced mean and clinically significant symptoms. Advantages were not demonstrated with 'therapeutic touch' or individualised activity. Aromatherapy and light therapy did not show clinical effectiveness. Training family carers in behavioural or cognitive interventions did not decrease severe agitation. The few studies reporting activities of daily living or quality-of-life outcomes found no improvement, even when agitation had improved. We identified two health economic studies. Costs of interventions which significantly impacted on agitation were activities, £80-696; music therapy, £13-27; sensory interventions, £3-527; and training paid caregivers in person-centred care or communication skills with or without behavioural management training and DCM, £31-339. Among the 11 interventions that were evaluated using the Cohen-Mansfield Agitation Inventory (CMAI), the incremental cost per unit reduction in CMAI score ranged from £162 to £3480 for activities, £4 for music therapy, £24 to £143 for sensory interventions, and £6 to £62 for training paid caregivers in person-centred care or communication skills with or without behavioural management training and DCM. Health and social care costs ranged from around £7000 over 3 months in people without clinically significant agitation symptoms to around £15,000 at the most severe agitation levels. There is some evidence that DEMQOL-Proxy-U scores decline with Neuropsychiatric Inventory agitation scores. A multicomponent intervention in participants with mild to moderate dementia had a positive monetary net benefit and a 82.2% probability of being cost-effective at a maximum willingness to pay for a quality-adjusted life-year of £20,000 and a 83.18% probability at a value of £30,000.

LIMITATIONS

Although there were some high-quality studies, there were only 33 reasonably sized (> 45 participants) randomised controlled trials, and lack of evidence means that we cannot comment on many interventions' effectiveness. There were no hospital studies and few studies in people's homes. More health economic data are needed.

CONCLUSIONS

Person-centred care, communication skills and DCM (all with supervision), sensory therapy activities, and structured music therapies reduce agitation in care-home dementia residents. Future interventions should change care home culture through staff training and permanently implement evidence-based treatments and evaluate health economics. There is a need for further work on interventions for agitation in people with dementia living in their own homes.

PROTOCOL REGISTRATION

The study was registered as PROSPERO no. CRD42011001370.

FUNDING

The National Institute for Health Research Health Technology Assessment programme.

Cardiovascular risk prediction models for people with severe mental illness: results from the prediction and management of cardiovascular risk in people with severe mental illnesses (PRIMROSE) research program

IMPORTANCE

People with severe mental illness (SMI), including schizophrenia and bipolar disorder, have excess rates of cardiovascular disease (CVD). Risk prediction models validated for the general population may not accurately estimate cardiovascular risk in this group.

OBJECTIVE

To develop and validate a risk model exclusive to predicting CVD events in people with SMI incorporating established cardiovascular risk factors and additional variables.

DESIGN, SETTING, AND PARTICIPANTS

We used anonymous/deidentified data collected between January 1, 1995, and December 31, 2010, from the Health Improvement Network (THIN) to conduct a primary care, prospective cohort and risk score development study in the United Kingdom. Participants included 38,824 people with a diagnosis of SMI (schizophrenia, bipolar disorder, or other nonorganic psychosis) aged 30 to 90 years. During a median follow-up of 5.6 years, 2324 CVD events (6.0%) occurred.

MAIN OUTCOMES AND MEASURES

Ten-year risk of the first cardiovascular event (myocardial infarction, angina pectoris, cerebrovascular accidents, or major coronary surgery). Predictors included age, sex, height, weight, systolic blood pressure, diabetes mellitus, smoking, body mass index (BMI), lipid profile, social deprivation, SMI diagnosis, prescriptions for antidepressants and antipsychotics, and reports of heavy alcohol use.

RESULTS

We developed 2 CVD risk prediction models for people with SMI: the PRIMROSE BMI model and the PRIMROSE lipid model. These models mutually excluded lipids and BMI. In terms of discrimination, from cross-validations for men, the PRIMROSE lipid model D statistic was 1.92 (95% CI, 1.80-2.03) and C statistic was 0.80 (95% CI, 0.76-0.83) compared with 1.74 (95% CI, 1.63-1.86) and 0.78 (95% CI, 0.75-0.82) for published Cox Framingham risk scores. The corresponding results in women were 1.87 (95% CI, 1.76-1.98) and 0.79 (95% CI, 0.76-0.82) for the PRIMROSE lipid model and 1.58 (95% CI, 1.48-1.68) and 0.77 (95% CI, 0.73-0.81) for the Cox Framingham model. Discrimination statistics for the PRIMROSE BMI model were comparable to those for the PRIMROSE lipid model. Calibration plots suggested that both PRIMROSE models were superior to the Cox Framingham models.

CONCLUSIONS AND RELEVANCE

The PRIMROSE BMI and lipid CVD risk prediction models performed better in SMI compared with models that include only established CVD risk factors. Further work on the clinical effectiveness and cost-effectiveness of the PRIMROSE models is needed to ascertain the best thresholds for offering CVD interventions.

Clinical effectiveness of a staff training intervention in mental health inpatient rehabilitation units designed to increase patients' engagement in activities (the Rehabilitation Effectiveness for Activities for Life [REAL] study): single-blind, cluster-randomised controlled trial

BACKGROUND

Mental health inpatient rehabilitation services focus on people with complex psychosis who have, for example, treatment-refractory symptoms, cognitive impairment, and severe negative symptoms, which impair functioning and require lengthy admission. Engagement in activities could lead to improvement in negative symptoms and function, but few trials have been done. We aimed to investigate the effectiveness of a staff training intervention to increase patients' engagement in activities.

METHODS

We did a single-blind, two-arm, cluster-randomised controlled trial in 40 mental health inpatient rehabilitation units across England. Units were randomly allocated to either a manual-based staff training programme delivered by a small intervention team (intervention group, n=20) or standard care (control group, n=20). The primary outcome was patients' engagement in activities 12 months after randomisation, measured with the time use diary. With this measure, both the degree of engagement in an activity and its complexity are recorded four times a day for a week, rated on a scale of 0-4 for every period (maximum score of 112). Analysis was by intention-to-treat. Random-effects models were used to compare outcomes between study groups. Cost-effectiveness was assessed by combining service costs with the primary outcome. This study is registered with Current Controlled Trials (ISRCTN25898179).

FINDINGS

Patients' engagement in activities did not differ between study groups (coefficient 1·44, 95% CI -1·35 to 4·24). An extra £101 was needed to achieve a 1% increase in patients' engagement in activities with the study intervention.

INTERPRETATION

Our training intervention did not increase patients' engagement in activities after 12 months of follow-up. This failure could be attributable to inadequate implementation of the intervention, a high turnover of patients in the intervention units, competing priorities on staff time, high levels of patients' morbidity, and ceiling effects because of the high quality of standard care delivered. Further studies are needed to identify interventions that can improve outcomes for people with severe and complex psychosis.

FUNDING

National Institute for Health Research.

Non-pharmacological interventions for agitation in dementia: systematic review of randomised controlled trials

BACKGROUND

Agitation in dementia is common, persistent and distressing and can lead to care breakdown. Medication is often ineffective and harmful.

AIMS

To systematically review randomised controlled trial evidence regarding non-pharmacological interventions. Method We reviewed 33 studies fitting predetermined criteria, assessed their validity and calculated standardised effect sizes (SES).

RESULTS

Person-centred care, communication skills training and adapted dementia care mapping decreased symptomatic and severe agitation in care homes immediately (SES range 0.3-1.8) and for up to 6 months afterwards (SES range 0.2-2.2). Activities and music therapy by protocol (SES range 0.5-0.6) decreased overall agitation and sensory intervention decreased clinically significant agitation immediately. Aromatherapy and light therapy did not demonstrate efficacy.

CONCLUSIONS

There are evidence-based strategies for care homes. Future interventions should focus on consistent and long-term implementation through staff training. Further research is needed for people living in their own homes.

A novel clinical risk prediction model for sudden cardiac death in hypertrophic cardiomyopathy (HCM risk-SCD)

AIMS

Hypertrophic cardiomyopathy (HCM) is a leading cause of sudden cardiac death (SCD) in young adults. Current risk algorithms provide only a crude estimate of risk and fail to account for the different effect size of individual risk factors. The aim of this study was to develop and validate a new SCD risk prediction model that provides individualized risk estimates.

METHODS AND RESULTS

The prognostic model was derived from a retrospective, multi-centre longitudinal cohort study. The model was developed from the entire data set using the Cox proportional hazards model and internally validated using bootstrapping. The cohort consisted of 3675 consecutive patients from six centres. During a follow-up period of 24 313 patient-years (median 5.7 years), 198 patients (5%) died suddenly or had an appropriate implantable cardioverter defibrillator (ICD) shock. Of eight pre-specified predictors, age, maximal left ventricular wall thickness, left atrial diameter, left ventricular outflow tract gradient, family history of SCD, non-sustained ventricular tachycardia, and unexplained syncope were associated with SCD/appropriate ICD shock at the 15% significance level. These predictors were included in the final model to estimate individual probabilities of SCD at 5 years. The calibration slope was 0.91 (95% CI: 0.74, 1.08), C-index was 0.70 (95% CI: 0.68, 0.72), and D-statistic was 1.07 (95% CI: 0.81, 1.32). For every 16 ICDs implanted in patients with ≥4% 5-year SCD risk, potentially 1 patient will be saved from SCD at 5 years. A second model with the data set split into independent development and validation cohorts had very similar estimates of coefficients and performance when externally validated.

CONCLUSION

This is the first validated SCD risk prediction model for patients with HCM and provides accurate individualized estimates for the probability of SCD using readily collected clinical parameters.

Study protocol: cluster randomised controlled trial to assess the clinical and cost effectiveness of a staff training intervention in inpatient mental health rehabilitation units in increasing service users' engagement in activities

BACKGROUND

This study focuses on people with complex and severe mental health problems who require inpatient rehabilitation. The majority have a diagnosis of schizophrenia whose recovery has been delayed due to non-response to first-line treatments, cognitive impairment, negative symptoms and co-existing problems such as substance misuse. These problems contribute to major impairments in social and everyday functioning necessitating lengthy admissions and high support needs on discharge to the community. Engagement in structured activities reduces negative symptoms of psychosis and may lead to improvement in function, but no trials have been conducted to test the efficacy of interventions that aim to achieve this.

METHODS/DESIGN

This study aims to investigate the clinical and cost-effectiveness of a staff training intervention to increase service users' engagement in activities. This is a single-blind, two-arm cluster randomised controlled trial involving 40 inpatient mental health rehabilitation units across England. Units are randomised on an equal basis to receive either standard care or a "hands-on", manualised staff training programme comprising three distinct phases (predisposing, enabling and reinforcing) delivered by a small team of psychiatrists, occupational therapists, service users and activity workers. The primary outcome is service user engagement in activities 12 months after randomisation, assessed using a standardised measure. Secondary outcomes include social functioning and costs and cost-effectiveness of care.

DISCUSSION

The study will provide much needed evidence for a practical staff training intervention that has potential to improve service user functioning, reducing the need for hospital treatment and supporting successful community discharge. The trial is registered with Current Controlled Trials (Ref ISRCTN25898179).

Service quality and clinical outcomes: an example from mental health rehabilitation services in England

BACKGROUND

Current health policy assumes better quality services lead to better outcomes.

AIMS

To investigate the relationship between quality of mental health rehabilitation services in England, local deprivation, service user characteristics and clinical outcomes.

METHOD

Standardised tools were used to assess the quality of mental health rehabilitation units and service users' autonomy, quality of life, experiences of care and ratings of the therapeutic milieu. Multiple level modelling investigated relationships between service quality, service user characteristics and outcomes.

RESULTS

A total of 52/60 (87%) National Health Service trusts participated, comprising 133 units and 739 service users. All aspects of service quality were positively associated with service users' autonomy, experiences of care and therapeutic milieu, but there was no association with quality of life.

CONCLUSIONS

Quality of care is linked to better clinical outcomes in people with complex and longer-term mental health problems. Thus, investing in quality is likely to show real clinical gains.

Endobronchial ultrasound-guided transbronchial needle aspiration prevents mediastinoscopies in the diagnosis of isolated mediastinal lymphadenopathy: a prospective trial

RATIONALE

Patients with isolated mediastinal lymphadenopathy (IML) are a common presentation to physicians, and mediastinoscopy is traditionally considered the "gold standard" investigation when a pathological diagnosis is required. Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is established as an alternative to mediastinoscopy in patients with lung cancer.

OBJECTIVE

To determine the efficacy and health care costs of EBUS-TBNA as an alternative initial investigation to mediastinoscopy in patients with isolated IML.

METHODS

Prospective multicenter single-arm clinical trial of 77 consecutive patients with IML from 5 centers between April 2009 and March 2011. All patients underwent EBUS-TBNA. If EBUS-TBNA did not provide a diagnosis, then participants underwent mediastinoscopy.

MEASUREMENTS AND MAIN RESULTS

EBUS-TBNA prevented 87% of mediastinoscopies (95% confidence interval [CI], 77-94%; P < 0.001) but failed to provide a diagnosis in 10 patients (13%), all of whom underwent mediastinoscopy. The sensitivity and negative predictive value of EBUS-TBNA in patients with IML were 92% (95% CI, 83-95%) and 40% (95% CI, 12-74%), respectively. One patient developed a lower respiratory tract infection after EBUS-TBNA, requiring inpatient admission. The cost of the EBUS-TBNA procedure per patient was £1,382 ($2,190). The mean cost of the EBUS-TBNA strategy was £1,892 ($2,998) per patient, whereas a strategy of mediastinoscopy alone was significantly more costly at £3,228 ($5,115) per patient (P < 0.001). The EBUS-TBNA strategy is less costly than mediastinoscopy if the cost per EBUS-TBNA procedure is less than £2,718 ($4,307) per patient.

CONCLUSIONS

EBUS-TBNA is a safe, highly sensitive, and cost-saving initial investigation in patients with IML. Clinical trial registered with ClinicalTrials.gov (NCT00932854).

Cerebral magnetic resonance biomarkers in neonatal encephalopathy: a meta-analysis

OBJECTIVE

Accurate prediction of neurodevelopmental outcome in neonatal encephalopathy (NE) is important for clinical management and to evaluate neuroprotective therapies. We undertook a meta-analysis of the prognostic accuracy of cerebral magnetic resonance (MR) biomarkers in infants with neonatal encephalopathy.

METHODS

We reviewed all studies that compared an MR biomarker performed during the neonatal period with neurodevelopmental outcome at > or =1 year. We followed standard methods recommended by the Cochrane Diagnostic Accuracy Method group and used a random-effects model for meta-analysis. Summary receiver operating characteristic curves and forest plots of each MR biomarker were calculated. chi(2) tests examined heterogeneity.

RESULTS

Thirty-two studies (860 infants with NE) were included in the meta-analysis. For predicting adverse outcome, conventional MRI during the neonatal period (days 1-30) had a pooled sensitivity of 91% (95% confidence interval [CI]: 87%-94%) and specificity of 51% (95% CI: 45%-58%). Late MRI (days 8-30) had higher sensitivity but lower specificity than early MRI (days 1-7). Proton MR spectroscopy deep gray matter lactate/N-acetyl aspartate (Lac/NAA) peak-area ratio (days 1-30) had 82% overall pooled sensitivity (95% CI: 74%-89%) and 95% specificity (95% CI: 88%-99%). On common study analysis, Lac/NAA had better diagnostic accuracy than conventional MRI performed at any time during neonatal period. The discriminatory powers of the posterior limb of internal capsule sign and brain-water apparent diffusion coefficient were poor.

CONCLUSIONS

Deep gray matter Lac/NAA is the most accurate quantitative MR biomarker within the neonatal period for prediction of neurodevelopmental outcome after NE. Lac/NAA may be useful in early clinical management decisions and counseling parents and as a surrogate end point in clinical trials that evaluate novel neuroprotective therapies.

Factors affecting self-efficacy and pain intensity in patients with chronic musculoskeletal pain seen in a specialist rheumatology pain clinic

OBJECTIVE

Chronic musculoskeletal pain is a very common and costly health problem. Patients presenting to rheumatology clinics with chronic pain can be difficult to manage. We studied 354 patients referred to a rheumatology chronic pain clinic over 5 yrs to identify factors affecting their self-efficacy and intensity of pain.

METHODS

We collected data for each patient, covering demographic and psychosocial factors, characteristics of their pain and previous treatment. We measured self-efficacy using a validated questionnaire, and pain intensity (PI) on an NRS. We performed multiple regression analysis to determine as to which factors were independently associated with these outcomes.

RESULTS

Despite extensive previous investigations and treatment, these patients had low self-efficacy [median = 26.5, interquartile range (IQR) 15-38, best possible = 60] and high PI scores (median = 7, worst possible = 10, IQR 5-9). Low self-efficacy was most clearly associated with depressive symptoms and not being employed. PI was most clearly associated with depressive symptoms, extensive pain and lower level of education.

CONCLUSION

Community-based studies suggest psychosocial factors are very important in determining outcomes in patients with chronic pain. This study suggests that the same is true in patients referred to rheumatologists due to chronic musculoskeletal pain and that these factors-particularly depressive symptoms and not being employed-are more important than site or duration of pain in those patients.

A model based on age, sex, and morbidity to explain variation in UK general practice prescribing: cohort study

OBJECTIVE

To examine whether patient level morbidity based measure of clinical case mix explains variations in prescribing in general practice.

DESIGN

Retrospective study of a cohort of patients followed for one year.

SETTING

UK General Practice Research Database.

PARTICIPANTS

129 general practices, with a total list size of 1 032 072.

MAIN OUTCOME MEASURES

Each patient was assigned a morbidity group on the bases of diagnoses, age, and sex using the Johns Hopkins adjusted clinical group case mix system. Multilevel regression models were used to explain variability in prescribing, with age, sex, and morbidity as predictors.

RESULTS

The median number of prescriptions issued annually to a patient is 2 (90% range 0 to 18). The number of prescriptions issued to a patient increases with age and morbidity. Age and sex explained only 10% of the total variation in prescribing compared with 80% after including morbidity. When variation in prescribing was split between practices and within practices, most of the variation was at the practice level. Morbidity explained both variations well.

CONCLUSIONS

Inclusion of a diagnosis based patient morbidity measure in prescribing models can explain a large amount of variability, both between practices and within practices. The use of patient based case mix systems may prove useful in allocation of budgets and therefore should be investigated further when examining prescribing patterns in general practices in the UK, particularly for specific therapeutic areas.

A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome

Risk models that aim to predict the future course and outcome of disease processes are increasingly used in health research, and it is important that they are accurate and reliable. Most of these risk models are fitted using routinely collected data in hospitals or general practices. Clinical outcomes such as short-term mortality will be near-complete, but many of the predictors may have missing values. A common approach to dealing with this is to perform a complete-case analysis. However, this may lead to overfitted models and biased estimates if entire patient subgroups are excluded. The aim of this paper is to investigate a number of methods for imputing missing data to evaluate their effect on risk model estimation and the reliability of the predictions. Multiple imputation methods, including hotdecking and multiple imputation by chained equations (MICE), were investigated along with several single imputation methods. A large national cardiac surgery database was used to create simulated yet realistic datasets. The results suggest that complete case analysis may produce unreliable risk predictions and should be avoided. Conditional mean imputation performed well in our scenario, but may not be appropriate if using variable selection methods. MICE was amongst the best performing multiple imputation methods with regards to the quality of the predictions. Additionally, it produced the least biased estimates, with good coverage, and hence is recommended for use in practice.

Association of age, sex and deprivation with quality indicators for diabetes: population-based cross sectional survey in primary care

OBJECTIVES

To determine the quality of diabetes management in primary care after the publication of the National Service Framework and examine the impact of age, gender and deprivation on the achievement of established quality indicators.

DESIGN

Population-based cross sectional survey using electronic general practice records carried out between June-October 2003.

SETTING

Thirty-four practices in Wandsworth, South-West London, UK.

PARTICIPANTS

6035 adult patients (> or =18 years) with diabetes from a total registered population of 201,572 patients.

INTERVENTIONS

None.

MAIN OUTCOME MEASURES

Success rates for the diabetes quality indicators within the General Medical Services contract for general practitioners.

RESULTS

We identified large variations in diabetes management between general practitioner practices with poorer recording of quality care in younger patients (18-44 years). In addition, younger patients had a worse cholesterol and glycaemia profile, although hypertension was more common in older patients. Gender and deprivation did not appear to be important determinants of the quality of care received.

CONCLUSIONS

There are large variations in diabetes management between general practitioner practices, with care seemingly worse for younger adults. Longitudinal studies are required to determine whether current UK quality improvement initiatives have been successful in attenuating existing variations in care and treatment outcomes.

Constructing intervals for the intracluster correlation coefficient using Bayesian modelling, and application in cluster randomized trials

Studies in health research are commonly carried out in clustered settings, where the individual response data are correlated within clusters. Estimation and modelling of the extent of between-cluster variation contributes to understanding of the current study and to design of future studies. It is common to express between-cluster variation as an intracluster correlation coefficient (ICC), since this measure is directly comparable across outcomes. ICCs are generally reported unaccompanied by confidence intervals. In this paper, we describe a Bayesian modelling approach to interval estimation of the ICC. The flexibility of this framework allows useful extensions which are not easily available in existing methods, for example assumptions other than Normality for continuous outcome data, adjustment for individual-level covariates and simultaneous interval estimation of several ICCs. There is also the opportunity to incorporate prior beliefs on likely values of the ICC. The methods are exemplified using data from a cluster randomized trial.

Meta-analysis of Studies of Alcohol and Breast Cancer with Consideration of the Methodological Issues

OBJECTIVE

To give an up-to-date assessment of the association of alcohol with female breast cancer, addressing methodological issues and shortfalls in previous overviews.

METHODS

Meta-analysis of studies (any language) providing original data on incidence of first primary breast cancer and alcohol. Two reviewers independently extracted data. Study quality assessed by objective criteria including degree of control for confounding; funnel plots examined for publication bias; meta-regression techniques to explore heterogeneity. Risks associated with drinking versus not drinking and dose-response not constrained through the origin estimated using random effects methods.

RESULTS

Ninety-eight unique studies were included, involving 75,728 and 60,653 cases in drinker versus non-drinker and dose-response analyses, respectively. Findings were robust to study design and analytic approaches in the meta-analyses. For studies judged high quality, controlled for appropriate confounders, excess risk associated with alcohol drinking was 22% (95% CI: 9-37%); each additional 10 g ethanol/day was associated with risk higher by 10% (95% CI: 5-15%). There was no evidence of publication bias. Risk did not differ significantly by beverage type or menopausal status. Estimated population attributable risks were 1.6 and 6.0% in USA and UK, respectively.

CONCLUSIONS

Taking account of shortcomings in the study base and methodological concerns, we confirm the alcohol-breast cancer association. We compared our results to those of an individual patient data analysis, with similar findings. We conclude that the association between alcohol and breast cancer may be causal.

Modelling Multivariate Outcomes in Hierarchical Data, with Application to Cluster Randomised Trials

In the cluster randomised study design, the data collected have a hierarchical structure and often include multivariate outcomes. We present a flexible modelling strategy that permits several normally distributed outcomes to be analysed simultaneously, in which intervention effects as well as individual-level and cluster-level between-outcome correlations are estimated. This is implemented in a Bayesian framework which has several advantages over a classical approach, for example in providing credible intervals for functions of model parameters and in allowing informative priors for the intracluster correlation coefficients. In order to declare such informative prior distributions, and fit models in which the between-outcome covariance matrices are constrained, priors on parameters within the covariance matrices are required. Careful specification is necessary however, in order to maintain non-negative definiteness and symmetry between the different outcomes. We propose a novel solution in the case of three multivariate outcomes, and present a modified existing approach and novel alternative for four or more outcomes. The methods are applied to an example of a cluster randomised trial in the prevention of coronary heart disease. The modelling strategy presented would also be useful in other situations involving hierarchical multivariate outcomes.

Quality research in healthcare: are researchers getting enough statistical support?

BACKGROUND

Reviews of peer-reviewed health studies have highlighted problems with their methodological quality. As published health studies form the basis of many clinical decisions including evaluation and provisions of health services, this has scientific and ethical implications. The lack of involvement of methodologists (defined as statisticians or quantitative epidemiologists) has been suggested as one key reason for this problem and this has been linked to the lack of access to methodologists. This issue was highlighted several years ago and it was suggested that more investments were needed from health care organisations and Universities to alleviate this problem.

METHODS

To assess the current level of methodological support available for health researchers in England, we surveyed the 25 National Health Services Trusts in England, that are the major recipients of the Department of Health's research and development (R&D) support funding.

RESULTS AND DISCUSSION

The survey shows that the earmarking of resources to provide appropriate methodological support to health researchers in these organisations is not widespread. Neither the level of R&D support funding received nor the volume of research undertaken by these organisations showed any association with the amount they spent in providing a central resource for methodological support for their researchers.

CONCLUSION

The promotion and delivery of high quality health research requires that organisations hosting health research and their academic partners put in place funding and systems to provide appropriate methodological support to ensure valid research findings. If resources are limited, health researchers may have to rely on short courses and/or a limited number of advisory sessions which may not always produce satisfactory results.

Generic, Simple Risk Stratification Model for Heart Valve Surgery

Background— Heart valve surgery has an associated in-hospital mortality rate of 4% to 8%. This study aims to develop a simple risk model to predict the risk of in-hospital mortality for patients undergoing heart valve surgery to provide information to patients and clinicians and to facilitate institutional comparisons.

Methods and Results— Data on 32 839 patients were obtained from the Society of Cardiothoracic Surgeons of Great Britain and Ireland on patients who underwent heart valve surgery between April 1995 and March 2003. Data from the first 5 years (n=16 679) were used to develop the model; its performance was evaluated on the remaining data (n=16 160). The risk model presented here is based on the combined data. The overall in-hospital mortality was 6.4%. The risk model included, in order of importance (all P <0.01), operative priority, age, renal failure, operation sequence, ejection fraction, concomitant tricuspid valve surgery, type of valve operation, concomitant CABG surgery, body mass index, preoperative arrhythmias, diabetes, gender, and hypertension. The risk model exhibited good predictive ability (Hosmer-Lemeshow test, P =0.78) and discriminated between high- and low-risk patients reasonably well (receiver-operating characteristics curve area, 0.77).

Conclusions— This is the first risk model that predicts in-hospital mortality for aortic and/or mitral heart valve patients with or without concomitant CABG. Based on a large national database of heart valve patients, this model has been evaluated successfully on patients who had valve surgery during a subsequent time period. It is simple to use, includes routinely collected variables, and provides a useful tool for patient advice and institutional comparisons.

Case-mix and variation in specialist referrals in general practice

BACKGROUND

The potential of a comprehensive measure of patient morbidity to explain variation in referrals to secondary care has not previously been examined in the UK.

AIM

To examine the relative role of age, sex and morbidity as defined by the Johns Hopkins ACG Case-Mix System in explaining variations in specialist referrals in general practice.

DESIGN OF STUDY

Retrospective study of a cohort of patients followed for 1 year.

SETTING

Two hundred and two general practices, with a total list size of 1,161,892, contributing data to the General Practice Research Database.

METHOD

Each patient was assigned an ACG and morbidity group, based on their diagnoses, age and sex. The variability in referrals explained by these factors was examined using multilevel logistic regression models by splitting it into variation between practices and variation between patients within practices.

RESULTS

The annual median (range) percentage of patients referred was 14.8% (range = 2.4-24.4%). The percentage of patients referred increased with age and morbidity. Morbidity explained 30.4% of the total variation in referrals (composed of variability between and within practices). Age and sex only explained 5.3% of the total variation. The variation attributable to practices was approximately 5%, thus most of the variation occurred within practices. Morbidity was also identified as a better predictor of referral compared to age and sex.

CONCLUSIONS

Morbidity explains almost six times more of the variation in general practice referrals than age and sex, although about two-thirds of the variation remains unexplained. Most of the unexplained variation is due to differences within rather than between practices. The amount of variability in referrals between practices may be less than implied by previous studies based on aggregate information. The implications are that any investigation of specialist referrals from general practice should be interpreted cautiously, even after adjustment for age, sex and morbidity.

Adjusting for case mix and social class in examining variation in home visits between practices

OBJECTIVES

The purpose of this study was to investigate whether adjusting for clinical case mix and social class explains more of the variation in home visits between general practices than adjusting for age and sex alone.

METHODS

The setting was 60 general practices in England and Wales taking part in the 1 year Fourth National Morbidity Survey. The participants comprised 349 505 patients who were registered with one of the participating general practices for at least 180 days, and who had at least one consultation during the period. The outcome measure is whether or not a patient received a home visit in that year. A clinical case mix category (morbidity class) based on 1 year's diagnostic information was assigned to each patient using the Johns Hopkins Adjusted Clinical Groups (ACG) Case Mix System. The social class measure was derived from occupation and employment status and is similar to that of the 1991 UK census. Variations in home visits between practices were examined using multilevel logistic regression models. The variability between practices before and after adjusting for clinical case mix and social class was estimated using the intracluster correlation coefficient (ICC).

RESULTS

The overall percentage of patients receiving a home visit over the 1 year study period was 17%, and this varied from 7 to 31% across the 60 practices. The percentage of the total variation in home visits attributable to differences between practices was 2.5% [95% confidence interval (CI) 1.4-3.2%] after adjusting for age and sex. This reduced to 1.6% (95% CI 1.1-2.4%) after taking into account morbidity class. The results were similar when social class was included instead of morbidity class. Morbidity and social class together reduced variation in home visits between practices to 1.5% (95% CI 1.1-2.2%).

CONCLUSIONS

Age, sex, social class and clinical case mix are strong determinants of home visits in the UK. Adjusting for morbidity and social class results in a small improvement in explaining the variability in home visits between practices compared with adjusting for age and sex alone. There is far more variation between patients within practices; however, it is not straightforward to examine the factors influencing this variation. In addition to morbidity and social class, there could also be other unmeasured factors such as varying patient demand for home visits, disability or differences in GP home visiting practice style that could influence the large within-practice variability observed in this study.

Cardiac surgery risk modeling for mortality: a review of current practice and suggestions for improvement

Risk models play a vital role in monitoring health care performances. Despite extensive research and widespread use of risk models in cardiac surgery, there are methodologic problems. We reviewed the methodology used for risk models for short-term mortality. The findings suggest that many risk models are developed in an ad hoc manner. Important aspects such as selection of risk factors, handling of missing values, and size of the data used for model development are not dealt with adequately. Methodologic details presented in publications are often sparse and unclear. Model development and validation processes are not always linked to the clinical aim of the model, which may affect their clinical validity. We make some suggestions in this review for improvement in methodology and reporting.

Off-Pump Coronary Artery Bypass (OPCAB) surgery reduces risk-stratified morbidity and mortality: a United Kingdom Multi-Center Comparative Analysis of Early Clinical Outcome

OBJECTIVE

Off-Pump Coronary Artery Bypass (OPCAB) surgery is gaining more popularity worldwide. The aim of this United Kingdom (UK) multi-center study was to assess the early clinical outcome of the OPCAB technique and perform a risk-stratified comparison with the conventional Coronary Artery Bypass Grafting (CABG) using the Cardio-Pulmonary Bypass (CPB) technique.

METHODS

Data were collected on 5,163 CPB patients from the database of the National Heart and Lung institute, Imperial College, University of London, and on 2,223 OPCAB patients from eight UK cardiac surgical centers, which run established OPCAB surgery programs. All patients had undergone primary isolated CABG for multi-vessel disease through a midline sternotomy approach, between January 1997 and April 2001. Postoperative morbidity and mortality were compared between the CPB and OPCAB patients after adjusting for case-mix. The mortality of the OPCAB patients was also compared, using risk stratification, to the mortality figures reported by the Society of Cardiothoracic Surgeons of Great Britain and Ireland (SCTS) based on 28,018 patients in the national database who were operated on between January 1996 and December 1999.

RESULTS

Morbidity and mortality were significantly lower in the OPCAB patients compared with the CPB patients and the UK national database of CABG patients, over the same period of time, after adjusting for case-mix.

CONCLUSIONS

This study demonstrates that risk stratified morbidity and mortality are significantly lower in OPCAB patients than CPB patients and patients in the UK national database.

Use of flexibility tests in the manufacturing process of 60° björk-shiley convexo-concave valves and the risk of outlet strut fracture

OBJECTIVES

Outlet strut fracture remains a concern for 30,000 patients living with a Björk-Shiley convexo-concave heart valve (Shiley, Inc, Irvine, Calif, a subsidiary of Pfizer, Inc). Previous studies (Netherlands and United Kingdom) investigating valve manufacturing aspects identified multiple performance of the hook deflection test as a risk factor for 60 degrees valves. The present study validated this finding using new data with a greater number of valves implanted worldwide. Risks of outlet strut fracture associated with other manufacturing aspects were also investigated.

METHODS

A matched case-control study design was used including 416 outlet strut fracture cases and 803 controls.

RESULTS

Analyses similar to that of the Dutch and United Kingdom studies produced odds ratios of 3.4 (95% confidence interval [CI]: 1.1-10.3) and 2.8 (95% CI: 1.1-7.3), respectively, for multiple hook deflection tests. Load deflection test, which replaced the hook deflection test, showed a statistically significant association with outlet strut fracture: odds ratio of 5.0 (95% CI: 2.1-11.8) and 6.2 (95% CI: 2.2-18.0) for single and multiple load deflection tests, respectively. An analysis where hook deflection tests were separated from load deflection tests showed significantly elevated odds ratios with performance of any type of flexibility test, and the highest odds ratio was observed with a combined performance of load and hook deflection tests.

CONCLUSIONS

Multiple hook deflection tests can now be considered for inclusion in the risk model used for guidelines on explant surgery to improve prediction of outlet strut fracture and provide patient reassurance. Load deflection tests and combined performance of hook and load deflection tests were found to be significant risk factors. No outlet strut fractures were reported for valves manufactured after March 1984 when the load deflection test was still in place. Examining manufacturing documents for these valves may identify new risk factors that could be responsible for the outlet strut fractures risk that remains unexplained to date.

An evaluation of existing risk stratification models as a tool for comparison of surgical performances for coronary artery bypass grafting between institutions

OBJECTIVE

Risk stratification systems are used in cardiac surgery to estimate mortality risk for individual patients and to compare surgical performance between institutions or surgeons. This study investigates the suitability of six existing risk stratification systems for these purposes.

METHODS

Data on 5471 patients who underwent isolated coronary artery bypass grafting at two UK cardiac centres between 1993 and 1999 were extracted from a prospective computerised clinical data base. Of these patients, 184 (3.3%) died in hospital. In-hospital mortality risk scores were calculated for each patient using the Parsonnet score, the EuroSCORE, the ACC/AHA score and three UK Bayes models (old, new complex and new simple). The accuracy for predicting mortality at an institutional level was assessed by comparing total observed and predicted mortality. The accuracy of the risk scores for predicting mortality for a patient was assessed by the Hosmer-Lemeshow test. The receiver operating characteristic (ROC) curve was used to evaluate how well a system ranks the patient with respect to their risk of mortality and can be useful for patient management.

RESULTS

Both EuroSCORE and the simple Bayes model were reasonably accurate at predicting overall mortality. However predictive accuracy at the patient level was poor for all systems, although EuroSCORE was accurate for low to medium risk patients. Discrimination was fair with the following ROC areas: Parsonnet 0.73, EuroSCORE 0.76, ACC/AHA system 0.76, old Bayes 0.77, complex Bayes 0.76, simple Bayes 0.76.

CONCLUSIONS

This study suggests that two of the scores may be useful in comparing institutions. None of the risk scores provide accurate risk estimates for individual patients in the two hospitals studied although EuroSCORE may have some utility for certain patients. All six systems perform moderately at ranking the patients and so may be useful for patient management. More results are needed from other institutions to confirm that the EuroSCORE and the simple Bayes model are suitable for institutional risk-adjusted comparisons.

Effectiveness of supervised physical therapy in the early period after arthroscopic partial meniscectomy

BACKGROUND AND PURPOSE

Controversy exists about the effectiveness of physical therapy after arthroscopic partial meniscectomy. This randomized controlled trial evaluated the effectiveness of supervised physical therapy with a home program versus a home program alone.

SUBJECTS

Eighty-four patients (86% males; overall mean age=39 years, SD=9, range=21-58; female mean age=39 years, SD=9, range=24-58; male mean age=40, SD=9, range=21-58) who underwent an uncomplicated arthroscopic partial meniscectomy participated.

METHODS

Subjects were randomly assigned to either a group who received 6 weeks of supervised physical therapy with a home program or a group who received only a home program. Blinded test sessions were conducted 5 and 50 days after surgery. Outcome measures were: (1) Hughston Clinic questionnaire, (2) Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) and EuroQol EQ-5D (EQ-5D) questionnaires, (3) number of days to return to work after surgery divided by the Factor Occupational Rating System score, (4) kinematic analysis of knee function during level walking and stair use, and (5) horizontal and vertical hops.

RESULTS

No differences between groups were found for any of the outcomes measured.

DISCUSSION AND CONCLUSION

The results indicate that the supervised physical therapy used in this study is not beneficial for patients in the early period after uncomplicated arthroscopic partial meniscectomy.

Importance of emergency identification schemes

BACKGROUND

Millions of people worldwide may have a hidden medical condition that could endanger their life in an emergency. These conditions may include cardiac conditions, severe allergies, or diabetes. Emergency identification schemes such as Medic Alert produce emblems that alert health care professionals to potential problems and can ensure appropriate and prompt treatment. This paper uses mechanical failure of the Björk-Shiley convexo-concave (BSCC) heart valve as an example of a hidden medical condition. These patients have been encouraged to carry information to alert staff in an emergency that they have a BSCC patient in their care and to be alert to the signs and symptoms of acute valve malfunction.

OBJECTIVE

To establish awareness and credibility of emergency identification schemes among emergency personnel and to assess if information on specific medical conditions would influence ambulance personnel regarding destination hospitals.

METHODS

Questionnaires were sent to senior staff (n=380) of accident and emergency (A&E) departments and operational directors of ambulance headquarters (n=39) throughout the United Kingdom. Hospitals were divided into regional divisions to assess differences in responses across regions.

RESULTS

The majority of respondents (99%) had heard of emergency identification schemes and felt that it was important for patients with special conditions to carry some form of identification. Nearly all ambulance respondents (97%) indicated it was routine to search for body worn emblems in contrast with only 71% of A & E staff. However, more than half of ambulance respondents (53.9%) stated information on emblems/cards would not influence their choice of destination hospital.

CONCLUSIONS

The importance of how information on pre-existing medical conditions can influence care, is highlighted by the BSCC valve issue, where immediate diagnosis is essential for patient survival. It is vital that all staff routinely search patients for this information and if necessary act upon the information provided.

Meta-analysis of ordinal outcomes using individual patient data

Meta-analyses are being undertaken in an increasing diversity of diseases and conditions, some of which involve outcomes measured on an ordered categorical scale. We consider methodology for undertaking a meta-analysis on individual patient data for an ordinal response. The approach is based on the proportional odds model, in which the treatment effect is represented by the log-odds ratio. A general framework is proposed for fixed and random effect models. Tests of the validity of the various assumptions made in the meta-analysis models, such as a global test of the assumption of proportional odds between treatments, are presented. The combination of studies with different definitions or numbers of response categories is discussed. The methods are illustrated on two data sets, in a classical framework using SAS and MLn and in a Bayesian framework using BUGS. The relative merits of the three software packages for such meta-analyses are discussed.

Meta-analysis of continuous outcome data from individual patients

Meta-analyses using individual patient data are becoming increasingly common and have several advantages over meta-analyses of summary statistics. We explore the use of multilevel or hierarchical models for the meta-analysis of continuous individual patient outcome data from clinical trials. A general framework is developed which encompasses traditional meta-analysis, as well as meta-regression and the inclusion of patient-level covariates for investigation of heterogeneity. Unexplained variation in treatment differences between trials is considered as random. We focus on models with fixed trial effects, although an extension to a random effect for trial is described. The methods are illustrated on an example in Alzheimer's disease in a classical framework using SAS PROC MIXED and MLwiN, and in a Bayesian framework using BUGS. Relative merits of the three software packages for such meta-analyses are discussed, as are the assessment of model assumptions and extensions to incorporate more than two treatments.

Outlet strut fracture of Björk-Shiley convexo concave heart valves: the UK cohort study

OBJECTIVE

To investigate the risk of outlet strut fracture (OSF) in Björk-Shiley convexo concave (BSCC) valves in relation to patients' clinical and valve characteristics.

DESIGN

A cohort of 2977 patients with 3325 valves with a follow up of 18 years.

SETTING

38 cardiac implantation centres in the UK.

RESULTS

56 OSF events were reported with 43 occurring in mitral and 13 in aortic valves. The overall OSF rate was 0.17%/year. No dominant clinical factor of risk was found, but multiple regression analysis identified age, body surface area, valve size, shop order fracture rate, and manufacturing period as risk factors for OSF. A 4% (95% confidence interval (CI) 2% to 6%) decrease in the risk of OSF was observed for each advancing year of age and a fivefold (95% CI 2 to 13) increase in risk for a 0.5 m(2) increase in body surface area. The association between the risk of OSF and valve size was not constant over time. Excess risks among 31 mm and 33 mm sizes (mainly mitral valves) decreased over time while that for 23 mm (almost all aortic valves) increased. The risk of OSF increased by 40% (95% CI 20% to 50%) for a unit increase in the fracture rate of other valves in the same batch. For valves manufactured during 1981 to 1984 the risk of OSF was 4 (95% CI 2 to 12) times greater than for valves manufactured before 1981.

CONCLUSIONS

The OSF rates for 60 degrees BSCC valves observed in the UK are the highest among all monitored populations. The changing patterns of mitral and aortic valve OSF rates over time observed in this study have not been identified previously and highlight the need for continued monitoring of patients with the BSCC valve.

Outlet strut fracture of Björk-Shiley convexo-concave valves: can valve-manufacturing characteristics explain the risk?

BACKGROUND

Björk-Shiley 60 degrees convexo-concave prosthetic heart valves (Shiley, Inc, Irvine, Calif, a subsidiary of Pfizer, Inc) continue to be a concern for approximately 35,000 nonexplanted patients worldwide, with approximately 600 events reported to the manufacturer to date. Fractures of the outlet struts of the valves began to appear in the early 1980s and have continued to the present, but their causes are only partially understood.

METHODS

A matched case-control study was conducted evaluating manufacturing records for 52 valves with outlet strut fractures and 248 control subjects matched for age at implantation, valve size, and valve position.

RESULTS

In addition to the risk factors recognized as determinants of outlet strut fracture, the United Kingdom case-control study has observed 7- to 9-fold increased risk with performance of multiple hook deflection tests. This test was performed more than once, usually after rework on the valve. Six valves in this study underwent multiple hook deflection tests, of which 4 experienced an outlet strut fracture. Cracks and further rework were noted for these valves. Significant associations were also observed between outlet strut fracture and disc-to-strut gap measurements taken before the attachment of the sewing ring.

CONCLUSIONS

It is our view that a combination of factors related to valve design, manufacturing process, and patient characteristics are responsible for outlet strut fractures of Björk-Shiley convexo-concave valves. Multiple hook deflection tests have emerged as a potential new risk factor for outlet strut fracture in both The Netherlands and the United Kingdom. This factor appears to be correlated with the presence of other abnormalities. A further study is needed to investigate the factors correlated with multiple hook deflection tests. On confirmation of risk, the presence of multiple hook deflection tests may be added to equations, quantifying the risk of outlet strut fracture for comparison against risk of mortality and serious morbidity from explant operations.

Risks of fracture of Björk-Shiley 60 degree convexo-concave prosthetic heart valves: long-term cohort follow up in the UK, Netherlands and USA

BACKGROUND AND AIM OF THE STUDY

Approximately 82,000 Björk-Shiley convexo-concave (BSCC) 60 degree prosthetic heart valves were implanted in patients worldwide between 1979 and 1986. Outlet strut fractures (OSF) of some of the valves were first reported shortly after their introduction. Here, the determinants of OSF are examined, and the between-country variation and long-term risk are assessed.

METHODS

Cohorts of patients in the UK, Netherlands and USA with 15,770 BSCC 60 degree heart valves were followed up to 18 years for the occurrence of OSF.

RESULTS

Crude rates of OSF were highest in the UK (0.18% per year), intermediate in the Netherlands (0.13%), and lowest in the USA (0.06%), although risk factor adjustment reduced the inter-country differences. Furthermore, in the UK and Netherlands, OSF rates (particularly for mitral valves) declined with time since implantation, and between-country differences were considerably diminished 10 or more years post implantation. The risk of OSF decreased steadily with advancing patient age. Fracture rates were lower among women than men, and also varied significantly with valve size and position and OSF status of other valves in the same shoporder.

CONCLUSION

This long-term follow up of BSCC 60 degree heart valve patients indicates that risk factors for valve fracture are generally similar in the UK, Netherlands and USA. It also identifies a strong association between fracture risk and age, newly reveals gender-related differences, and shows that the risk of valve fracture persisted, albeit at a reduced rate, into the 1990s.

Bayesian methods of analysis for cluster randomized trials with binary outcome data

We explore the potential of Bayesian hierarchical modelling for the analysis of cluster randomized trials with binary outcome data, and apply the methods to a trial randomized by general practice. An approximate relationship is derived between the intracluster correlation coefficient (ICC) and the between-cluster variance used in a hierarchical logistic regression model. By constructing an informative prior for the ICC on the basis of available information, we are thus able implicitly to specify an informative prior for the between-cluster variance. The approach also provides us with a credible interval for the ICC for binary outcome data. Several approaches to constructing informative priors from empirical ICC values are described. We investigate the sensitivity of results to the prior specified and find that the estimate of intervention effect changes very little in this data set, while its interval estimate is more sensitive. The Bayesian approach allows us to assume distributions other than normality for the random effects used to model the clustering. This enables us to gain insight into the robustness of our parameter estimates to the classical normality assumption. In a model with a more complex variance structure, Bayesian methods can provide credible intervals for a difference between two variance components, in order for example to investigate whether the effect of intervention varies across clusters. We compare our results with those obtained from classical estimation, discuss the relative merits of the Bayesian framework, and conclude that the flexibility of the Bayesian approach offers some substantial advantages, although selection of prior distributions is not straightforward.

A multilevel model framework for meta-analysis of clinical trials with binary outcomes

In this paper we explore the potential of multilevel models for meta-analysis of trials with binary outcomes for both summary data, such as log-odds ratios, and individual patient data. Conventional fixed effect and random effects models are put into a multilevel model framework, which provides maximum likelihood or restricted maximum likelihood estimation. To exemplify the methods, we use the results from 22 trials to prevent respiratory tract infections; we also make comparisons with a second example data set comprising fewer trials. Within summary data methods, confidence intervals for the overall treatment effect and for the between-trial variance may be derived from likelihood based methods or a parametric bootstrap as well as from Wald methods; the bootstrap intervals are preferred because they relax the assumptions required by the other two methods. When modelling individual patient data, a bias corrected bootstrap may be used to provide unbiased estimation and correctly located confidence intervals; this method is particularly valuable for the between-trial variance. The trial effects may be modelled as either fixed or random within individual data models, and we discuss the corresponding assumptions and implications. If random trial effects are used, the covariance between these and the random treatment effects should be included; the resulting model is equivalent to a bivariate approach to meta-analysis. Having implemented these techniques, the flexibility of multilevel modelling may be exploited in facilitating extensions to standard meta-analysis methods.

Analysis of a cluster randomized trial with binary outcome data using a multi-level model

The use of multi-level logistic regression models was explored for the analysis of data from a cluster randomized trial investigating whether a training programme for general practitioners' reception staff could improve women's attendance at breast screening. Twenty-six general practices were randomized with women nested within them, requiring a two-level model which allowed for between-practice variability. Comparisons were made with fixed effect (FE) and random effects (RE) cluster summary statistic methods, ordinary logistic regression and a marginal model based on generalized estimating equations with robust variance estimates. An FE summary statistic method and ordinary logistic regression considerably understated the variance of the intervention effect, thus overstating its statistical significance. The marginal model produced a higher statistical significance for the intervention effect compared to that obtained from the RE summary statistic method and the multi-level model. Because there was only a moderate number of practices and these had unbalanced cluster sizes, reliable asymptotic properties for the robust standard errors used in the marginal model may not have been achieved. While the RE summary statistic method cannot handle multiple covariates easily, marginal and multi-level models can do so. In contrast to multi-level models however, marginal models do not provide direct estimates of variance components, but treat these as nuisance parameters. Estimates of the variance components were of particular interest in this example. Additionally, parametric bootstrap methods within the multi-level model framework provide confidence intervals for these variance components, as well as a confidence interval for the effect of intervention which allows for the imprecision in the estimated variance components. The assumption of normality of the random effects can be checked, and the models extended to investigate multiple sources of variability.

Analysing repeated measurements data: a practical comparison of methods

A variety of methods are available for analysing repeated measurements data where the outcome is continuous. However, there is little information on how established methods, such as summary statistics and repeated measures analysis of variance (RMAOV), compare in practice with methods that have become available to applied statisticians more recently, such as marginal models (based on generalized estimating equation methodology) and multilevel models (that is, hierarchical random effects models). The aim of this paper is to exemplify the use of these methods, and directly compare their results by application to a clinical trial data set. The focus is on practical aspects rather than technical issues. The data considered were taken from a clinical trial of treatments for asthma in 240 children, in which a baseline and four post-randomization measurements of outcomes were taken. The simplicity of the method of summary statistics using the post-randomization mean of observations provided a useful initial analysis. However, fixed time effects or treatment-time interactions cannot be included in such an analysis, and choice of appropriate weighting when there is substantial missing data is problematic. RMAOV, marginal models and multilevel models generally provided similar estimates and standard errors for the treatment effects, although in one example with a relatively complex variance structure the marginal model produced less efficient estimates. Two advantages of multilevel models are that they provide direct estimates of variance components which are often of interest in their own right, and that they can be naturally extended to handle multivariate outcomes.

Cancer mortality and morbidity among plutonium workers at the Sellafield plant of British Nuclear Fuels

The mortality of all 14 319 workers employed at the Sellafield plant of British Nuclear Fuels between 1947 and 1975 was studied up to the end of 1992, and cancer incidence was examined from 1971 to 1986, in relation to their exposures to plutonium and to external radiation. The cancer mortality rate was 5% lower than that of England and Wales and 3% less than that of Cumbria. The significant excesses of deaths from cancer of the pleura and thyroid found in an earlier study persist with further follow-up (14 observed, 4.0 expected for pleura; 6 observed, 2.2 expected for thyroid). All of the deaths from pleural cancer were among radiation workers. For neither site was there a significant association between the risk of the cancer and accumulated radiation dose. There were significant deficits of deaths from cancers of mouth and pharynx, liver and gall bladder, and larynx and leukaemia when compared with the national rates. Among all radiation workers, there was a significant positive association between accumulated external radiation dose and mortality from cancers of ill-defined and secondary sites (10-year lag, P = 0.04), leukaemia (no lag, P = 0.03; 2-year lag, P = 0.05), multiple myeloma (20-year lag, P = 0.02), all lymphatic and haematopoietic cancers (20-year lag, P= 0.03) and all causes of death combined (20-year lag, P= 0.008). Among plutonium workers, there were significant excesses of deaths from cancer of the breast (6 observed, 2.6 expected) and ill-defined and secondary cancers (29 observed, 20.1 expected). No significant positive trends were observed between the risk of deaths from cancers of any specific site, or all cancers combined, and cumulative plutonium and external radiation doses. For no cancer site was there a significant excess of cancer registrations compared with rates for England and Wales. Analysis of trends in cancer incidence showed significant increases in risk with cumulative plutonium plus external radiation doses for all lymphatic and haematopoietic neoplasms for 0-, 10- and 20-year lag periods. Taken as a whole, our findings do not suggest that workers at Sellafield who have been exposed to plutonium are at an overall significantly increased risk of cancer compared with other radiation workers.

Cancer mortality in relation to monitoring for radionuclide exposure in three UK nuclear industry workforces

Cancer mortality in 40,761 employees of three UK nuclear industry facilities who had been monitored for external radiation exposure was examined according to whether they had also been monitored for possible internal exposure to tritium, plutonium or other radionuclides (uranium, polonium, actinium or other unspecified). Death rates from cancer were compared both with national rates and with rates in radiation workers not monitored for exposure to any radionuclides. Among workers monitored for tritium exposure, overall cancer mortality was significantly below national rates [standardized mortality ratio (SMR) = 83, 165 deaths; 2P = 0.02] and none of the cancer-specific death rates was significantly above either the national average or rates in non-monitored workers. Although the overall death rate from cancer in workers monitored for plutonium exposure was also significantly low relative to national rates (SMR = 89, 581 deaths; 2P = 0.005), mortality from pleural cancer was significantly raised (SMR = 357, nine deaths; 2P = 0.002); none of the rates differed significantly from those of non-monitored workers. Workers monitored for radionuclides other than tritium or plutonium also had a death rate from all cancers combined that was below the national average (SMR = 86, 418 deaths; 2P = 0.002) but prostatic cancer mortality was raised both in relation to death rates in the general population (SMR = 153, 37 deaths; 2P = 0.02) and to death rates in radiation workers who had not been monitored for exposure to any radionuclide [rate ratio (RR) = 1.65; 2P = 0.03]. Mortality from cancer of the lung was also significantly increased in workers monitored for other radionuclides compared with those of radiation workers not monitored for exposure to radionuclides (RR = 1.31, 164 deaths; 2P = 0.01). For cancers of the lung, prostate and all cancers combined, death rates in monitored workers were examined according to the timing and duration of monitoring for radionuclide exposure, with rates of radiation workers not monitored for any radionuclide forming the comparison group. In tritium-monitored workers, RRs for prostatic cancer varied significantly according to the number of years in which they were monitored (2P = 0.03). In workers monitored for plutonium exposure, RRs for all cancers combined increased with the number of years in which they were monitored (2P = 0.04) and with the number of years since first monitoring (2P = 0.0003). There was little suggestion of systematic variation in RRs for workers monitored for other radionuclides in relation to the timing or duration of monitoring, nor did it appear that their raised rates of cancer of the lung and prostate were explained by external radiation dose. These analyses of cancer mortality in relation to monitoring for radionuclide exposure reported in a large cohort of nuclear industry workers suggest that certain patterns of monitoring for some radionuclides may be associated with higher death rates from cancers of the lung, pleura, prostate and all cancers combined. Some of these findings may be due to chance. Moreover, because of the paucity of related data and lack of information about other possible exposures, such as whether plutonium workers are more likely to be exposed to asbestos, firm conclusions cannot be drawn at this stage. Further investigations of the relationship between radionuclide exposure and cancer in nuclear industry workers are needed.

Improving uptake of breast screening in multiethnic populations: a randomised controlled trial using practice reception staff to contact non-attenders

OBJECTIVES

To determine whether a two hour training programme for general practice reception staff could improve uptake in patients who had failed to attend for breast screening, and whether women from different ethnic groups benefited equally.

DESIGN

Controlled trial, randomised by general practice.

SETTING

Inner London borough of Newham.

SUBJECTS

2064 women aged 50-64 years who had failed to attend for breast screening. Women came from 26 of 37 eligible practices, 31% were white, 17% were Indian, 10% Pakistani, 14% black, 6% Bangladeshi, 1% Chinese, 4% were from other ethnic groups, and in 16% the ethnic group was not reported.

MAIN OUTCOME MEASURES

Attendance for breast screening in relation to ethnic group in women who had not taken up their original invitation.

RESULTS

Attendance in the intervention group was significantly better than in the control group (9% v 4%). The response was best in Indian women--it was 19% in the intervention group and 5% in the control group.

CONCLUSIONS

This simple, low cost intervention improved breast screening rates modestly. Improvement was greatest in Indian women--probably because many practice staff shared their cultural and linguistic background. This intervention could be effective as part of a multifaceted strategy to improve uptake in areas with low rates.

Second malignancy in patients with Hodgkin's disease treated at the Royal Marsden Hospital

Risk of second primary malignancy was assessed in follow-up to June 1991 of 1039 patients first treated for Hodgkin's disease at the Royal Marsden Hospital during 1963-91. A total of 77 second malignancies occurred. There were significantly raised risks of stomach [standardized incidence ratio (SIR)=4.0], lung (SIR=3.8), bone (SIR=26.5), soft tissue (SIR=16.9) and non-melanoma skin (SIR=3.9) cancers, non-Hodgkin's lymphoma (SIR=4.6), and acute and non-lymphocytic leukaemia (SIR=31.3), with a relative risk of 3.3 for all second cancers other than non-melanoma skin cancer. Solid cancer risk was raised to a similar extent in patients treated only with radiotherapy (SIR=2.6, P<0.001), only with chemotherapy (SIR=2.1, P=0.08) and with both (SIR=3.1, P<0.001). Leukaemia risk was raised only in those receiving chemotherapy, whether alone or with radiotherapy. The relative risk for solid cancers was much greater in patients who were younger at first treatment (trend P<0.001), whereas leukaemia risk was greatest for those first treated at ages 25-44. For solid cancers (P<0.001) but not leukaemia (P=0.05) there was a strong gradient of greater relative risks at younger attained ages. The relative risk of second cancers overall was 27.5 at ages under 25 and 2.0 at ages 55 and above. Leukaemia and solid cancer risks in patients treated with chlorambucil, vinblastine, procarbazine and prednisone (ChlVPP) were not significantly greater than those in patients treated with mustine, vincristine, procarbazine and prednisone (MOPP). Number of cycles of chemotherapy was significantly related to risk of leukaemia (P<0.001), and there was a trend in the same direction for solid cancers (P=0.07). The study adds to evidence that alkylating chemotherapy may increase the risk of solid cancers, and that ChlVPP does not provide a less carcinogenic alternative to MOPP chemotherapy. The very large relative risks found for solid cancers at young attained ages and in patients treated when young may have important implications as, in the long term, the majority of second malignancies after Hodgkin's disease are solid cancers. The risks of solid malignancies need clarification by larger collaborative epidemiological studies.

A parametric multistate model for the analysis of carcinogenicity experiments

A fully parametric multistate model is explored for the analysis of animal carcinogenicity experiments in which the time of tumour onset is not known. This model does not require assumptions about tumour lethality or cause of death judgements and can be fitted in the absence of sacrifice data. The model is constructed as a three-state model with simple parametric forms for the transition rates. Maximum likelihood methods are used to estimate the transition rates and different treatment groups are compared using likelihood ratio tests. Selection of an appropriate model and methods to assess the fit of the model are illustrated with data from animal experiments. Comparisons with standard methods are made.

Cancer mortality and morbidity among workers at the Sellafield plant of British Nuclear Fuels

The mortality of all 14,282 workers employed at the Sellafield plant of British Nuclear Fuels between 1947 and 1975 was studied up to the end of 1988 and cancer incidence was examined from 1971 to 1986. This updates a previous report on mortality only up to the end of 1983. Ninety-nine per cent of the workers were traced satisfactorily. Cancer mortality was 4% less than that of England and Wales [standardised mortality ratio (SMR) = 96; 95% confidence interval (CI) = 90,103] and the same as that of Cumbria (SMR = 100: Cl = 94,107). Cancer incidence was 10% less than that of England and Wales [standardised registration ratio (SRR) = 90; Cl = 83.97] and 18% less than that of Northern Region (SRR = 82; Cl = 75.88). Cancer mortality rates were significantly in excess of national rates for cancers of the pleura (nine observed, 2.6 expected; P = 0.001), thyroid (six observed, 1.8 expected; P = 0.01) and ill defined and secondary sites (53 observed, 39.2 expected; P = 0.02). There were significant deficits of cancers of the liver and gall bladder, larynx and lung. Among radiation workers there were significant positive correlations between accumulated radiation dose and mortality from cancers of ill-defined and secondary sites (10 year lag: P = 0.01) and for leukaemia (2 year lag: P = 0.009), but not for cancers of the pleura and thyroid cancer. Previous findings of such associations with multiple myeloma and bladder cancer were less strong. There was a significant excess of incident cases of cancer of the oesophagus (P = 0.01), but this was not associated with accumulated radiation dose. For cancers other than leukaemia, the dose-response risk estimates were below those of the adult atomic bomb survivors, but the 90% confidence interval included risks of zero and of 2-3 times higher. For leukaemia (12 deaths, excluding CLL), under an excess relative risk model, the risk estimate derived for the Sellafield workers was about four times higher than that for the adult atomic bomb survivors with a confidence interval ranging from a half to nearly 20 times that of the atomic bomb survivors. Overall, however, there was no excess of leukaemia among the workers compared with national rates.