The burden of musculoskeletal disorders in Belgium: a national population-based study

Background

According to WHO, approximately one in three people worldwide live with a chronic, painful musculoskeletal (MSK) disorder. Low back pain (LBP), neck pain (NKP), osteoarthritis (OST) and rheumatoid arthritis (RHE) are among the most disabling MSK disorders. According to the Global Burden of Disease (GBD) study, LBP was the leading global cause in terms of years lived with disability and OST showed an increase in prevalence and is predicted to be one of the leading future causes. Our study aimed to analyse the burden of these MSK disorders in Belgium, providing a summary of morbidity and mortality outcomes from 2013 to 2018.

Methods

Prevalence and disability-adjusted life years (DALY) were computed using data from the Belgian health interview surveys from 2013 and 2018, the INTEGO database (Belgian registration network for general practitioners) and GBD study 2019. Mortality data was retrieved from the Belgian statistical office for people dying from RHE. Following GBD methodology, LBP, NKP and OST were assumed to not generate any deaths.

Results

The prevalence of MSK disorders increased from 2013 to 2018 with OST being the disorder with the highest number of cases (1.7 million cases in 2018). The burden was higher in women and the gender disparities increased with age. Women died also more frequently due to RHE compared to men. In total MSK disorders contributed to 180,746 comorbidity-adjusted DALYs for female and 116,063 comorbidity-adjusted DALYs for men in 2018, with LBP being the largest contributor (140,031 DALY).

Conclusions

The burden of MSK disorders has increased over the years. In 2018, 2.5 million Belgians were affected by at least one MSK disorder that resulted in almost 300,000 DALY. Our study provides valuable information of a part of the health burden that is known to have a great impact on the total burden of disease but that is sometimes disregarded by public health institutions.

Key messages

• MSK disorders represent a major health problem in Belgium.

• Acting on risk factors associated to these disorders is crucial to mitigate their burden.

Direct and indirect costs attributable to musculoskeletal disorders in Belgium

Background

Within the European Union, musculoskeletal (MSK) disorders represent the most prevalent and costly work-related health problems affecting about 45 million workers. Since middle-aged people during their formative and peak income-earning years are predominantly affected, MSK disorders are the major contributors to the loss of productive life years in the workforce compared with other non-communicable diseases. This study aimed to summarize the average yearly economic impact of low back pain (LBP), neck pain (NKP), osteoarthritis (OST) and rheumatoid arthritis (RHE) in Belgium from 2013 to 2017.

Methods

Direct costs, measured by reimbursed expenditures for medical services and medications, were derived by the national health insurer. Indirect costs were computed by multiplying the mean number of days absent from work (derived by the Belgian health interview survey, as prevalence data) with the average gross daily wage. Multivariate regression models were used to explore the extent to which average yearly costs were associated with MSK disorders. The method of recycled predictions allowed to estimate the marginal effect of each MSK disorder on costs.

Results

25% of Belgian adults were affected by at least one MSK disorder that incurred on average to 1,524€ per capita. LBP was the most costly disorder (2,405€ per capita) followed by NKP (2,260€ per capita). In the working population, 15% had at least one MSK disorder with an average indirect cost of 3,083€ per capita. People with LBP were the only showing a significantly higher indirect cost compared to a population without LBP, with an adjusted cost per capita of 5,875€.

Conclusions

The adult Belgian population is largely affected by MSK disorders. Every year the total adjusted healthcare cost amounted to more than 3 billion Euros. Additionally, on average every year Belgium spends around 2 billion Euros for work absenteeism related to one of the MSK disorders.

Key messages

Validity of self-reported data to assess the prevalence of overweight, hypertension and cholesterol

Background

Accurate data on the risk factors of non-communicable diseases is essential to build evidence-based prevention programs. In Belgium, this is assessed through self-reported (SR) data from the Belgian health interview surveys (BHIS) in a wide population sample or through objective measures from small-scale surveys (such as the Belgian health examination survey [BHES]). It has been shown, however, that relying on SR data leads to a prevalence underestimation. The objective of this study is to assess the agreement between SR and measured overweight, hypertension and high cholesterol and to provide information to do a valid correction for measurement error.

Methods

The BHIS/BHES 2018 database was used (n = 1184). Kappa coefficient was used to assess the agreement between SR and measured hypertension (systolic BP > 140 mmHg, diastolic BP > 90 mmHg, or reported use of medication for hypertension) and ICC was used to assess the agreement between SR and measured BMI. SR high cholesterol was compared to a measured total serum cholesterol >190 mg/dl.

Results

Risk factor prevalence based on SR data is severely underestimated. The agreement between SR and measured data is high for BMI (ICC: 0.92), moderate for hypertension (Kappa: 0.49) and poor for cholesterol (Kappa: 0.05). Using SR data, 45% of the people with a measured hypertension and 22% of the people with a measured high cholesterol are detected. With regressions based on the SR risk factor, age, sex and education, the measured BMI and hypertension can be predicted with a good accuracy (BMI: R2: 87%, HBP: AUC: 86%). A lower accuracy is observed for the cholesterol model (AUC: 65%). Using predicted values instead of SR data yields higher estimates of people suffering from overweight (+8% relative increase), obesity (+12%), hypertension (+24%) and cholesterol (+36%).

Conclusions

Using SR data yields to an underestimation of the prevalence of obesity, hypertension and high cholesterol in Belgium

Key messages

Relying on SR data to assess the prevalence of overweight, hypertension and high cholesterol requires a correction for measurement error. Using the predicted values from regression models based on the SR risk factor, age, sex and education, yields higher estimates of people suffering from overweight, hypertension and cholesterol.

Outcome in patients perceived as receiving excessive care across different ethical climates: a prospective study in 68 intensive care units in Europe and the USA

Purpose

Whether the quality of the ethical climate in the intensive care unit (ICU) improves the identification of patients receiving excessive care and affects patient outcomes is unknown.

Methods

In this prospective observational study, perceptions of excessive care (PECs) by clinicians working in 68 ICUs in Europe and the USA were collected daily during a 28-day period. The quality of the ethical climate in the ICUs was assessed via a validated questionnaire. We compared the combined endpoint (death, not at home or poor quality of life at 1 year) of patients with PECs and the time from PECs until written treatment-limitation decisions (TLDs) and death across the four climates defined via cluster analysis.

Results

Of the 4747 eligible clinicians, 2992 (63%) evaluated the ethical climate in their ICU. Of the 321 and 623 patients not admitted for monitoring only in ICUs with a good (n = 12, 18%) and poor (n = 24, 35%) climate, 36 (11%) and 74 (12%), respectively were identified with PECs by at least two clinicians. Of the 35 and 71 identified patients with an available combined endpoint, 100% (95% CI 90.0–1.00) and 85.9% (75.4–92.0) (P = 0.02) attained that endpoint. The risk of death (HR 1.88, 95% CI 1.20–2.92) or receiving a written TLD (HR 2.32, CI 1.11–4.85) in patients with PECs by at least two clinicians was higher in ICUs with a good climate than in those with a poor one. The differences between ICUs with an average climate, with (n = 12, 18%) or without (n = 20, 29%) nursing involvement at the end of life, and ICUs with a poor climate were less obvious but still in favour of the former.

Conclusion

Enhancing the quality of the ethical climate in the ICU may improve both the identification of patients receiving excessive care and the decision-making process at the end of life.

Electronic supplementary material

The online version of this article (10.1007/s00134-018-5231-8) contains supplementary material, which is available to authorized users.

Does appropriate empiric antibiotic therapy modify intensive care unit-acquired Enterobacteriaceae bacteraemia mortality and discharge?

BACKGROUND

Conflicting results have been found regarding outcomes of intensive care unit (ICU)-acquired Enterobacteriaceae bacteraemia and the potentially modifying effect of appropriate empiric antibiotic therapy.

AIM

To evaluate these associations while adjusting for potential time-varying confounding using methods from the causal inference literature.

METHODS

Patients who stayed more than two days in two general ICUs in England between 2002 and 2006 were included in this cohort study. Marginal structural models with inverse probability weighting were used to estimate the mortality and discharge associated with Enterobacteriaceae bacteraemia and the impact of appropriate empiric antibiotic therapy on these outcomes.

FINDINGS

Among 3411 ICU admissions, 195 (5.7%) ICU-acquired Enterobacteriaceae bacteraemia cases occurred. Enterobacteriaceae bacteraemia was associated with an increased daily risk of ICU death [cause-specific hazard ratio (HR): 1.48; 95% confidence interval (CI): 1.10-1.99] and a reduced daily risk of ICU discharge (HR: 0.66; 95% CI: 0.54-0.80). Appropriate empiric antibiotic therapy did not significantly modify ICU mortality (HR: 1.08; 95% CI: 0.59-1.97) or discharge (HR: 0.91; 95% CI: 0.63-1.32).

CONCLUSION

ICU-acquired Enterobacteriaceae bacteraemia was associated with an increased daily risk of ICU mortality. Furthermore, the daily discharge rate was also lower after acquiring infection, even when adjusting for time-varying confounding using appropriate methodology. No evidence was found for a beneficial modifying effect of appropriate empiric antibiotic therapy on ICU mortality and discharge.

Causal mediation analysis with multiple mediators

In diverse fields of empirical research-including many in the biological sciences-attempts are made to decompose the effect of an exposure on an outcome into its effects via a number of different pathways. For example, we may wish to separate the effect of heavy alcohol consumption on systolic blood pressure (SBP) into effects via body mass index (BMI), via gamma-glutamyl transpeptidase (GGT), and via other pathways. Much progress has been made, mainly due to contributions from the field of causal inference, in understanding the precise nature of statistical estimands that capture such intuitive effects, the assumptions under which they can be identified, and statistical methods for doing so. These contributions have focused almost entirely on settings with a single mediator, or a set of mediators considered en bloc; in many applications, however, researchers attempt a much more ambitious decomposition into numerous path-specific effects through many mediators. In this article, we give counterfactual definitions of such path-specific estimands in settings with multiple mediators, when earlier mediators may affect later ones, showing that there are many ways in which decomposition can be done. We discuss the strong assumptions under which the effects are identified, suggesting a sensitivity analysis approach when a particular subset of the assumptions cannot be justified. These ideas are illustrated using data on alcohol consumption, SBP, BMI, and GGT from the Izhevsk Family Study. We aim to bridge the gap from "single mediator theory" to "multiple mediator practice," highlighting the ambitious nature of this endeavor and giving practical suggestions on how to proceed.

A retrospective study of the pregnancy, delivery and neonatal outcome in overweight versus normal weight women with polycystic ovary syndrome

STUDY QUESTION

Do overweight women with polycystic ovary syndrome (PCOS) have a higher risk of perinatal complications than normal weight women with PCOS?

SUMMARY ANSWER

Overweight women with PCOS with an ongoing singleton pregnancy have an increased risk of preterm birth as well as an increased risk of giving birth to a baby with a higher birthweight than normal weight women with PCOS.

WHAT IS KNOWN ALREADY

There is evidence that overweight (BMI > 25 kg/m²) has a negative influence on the prevalence of gestational diabetes mellitus and fetal macrosomia in women with PCOS.

STUDY DESIGN, SIZE, DURATION

We set up a retrospective comparative cohort study of 93 overweight (BMI ≥ 25 kg/m²) and 107 normal weight (BMI < 25 kg/m²) women with PCOS who were scheduled for fertility treatment between January 2000 and December 2009 and achieved a pregnancy as a result of a treatment cycle, or spontaneously before or between treatment cycles.

PARTICIPANTS/MATERIALS, SETTING, METHODS

All data (patient characteristics, medical information, pregnancy, delivery and neonatal outcome) were retrieved from patient medical files. All pregnancy, delivery and neonatal outcome parameters were adjusted for age and pre-pregnancy smoking behaviour. The neonatal outcome parameters were additionally adjusted for gestational age.

MAIN RESULTS AND THE ROLE OF CHANCE

The median BMI in the overweight and normal weight women was, respectively, 30.8 kg/m² [interquartile quartile range (IQR) 5.8] and 20.9 kg/m² (IQR 2.3) (P < 0.001). Baseline characteristics did not differ between groups, except for free testosterone and fasting insulin levels, which were higher, and sex hormone-binding globulin, which was lower, in overweight versus normal weight women (all P < 0.001). The time-to-pregnancy was significantly higher in the overweight group (P = 0.01). Multivariate analyses of the ongoing singleton pregnancies showed significantly more preterm births in overweight (10/61) versus normal weight (2/71) women [adjusted odds ratio 0.1, 95% confidence interval (CI) 0-0.6, P = 0.01]. The mean birthweight of newborns was significantly higher in overweight (3386 ± 663 g) than in normal weight (3251 ± 528 g) women (adjusted mean difference 259.4, 95% CI 83.4-435.4, P = 0.004).

LIMITATIONS, REASON FOR CAUTION

Our results only represent the pregnancy, delivery and neonatal outcome of ongoing singleton pregnancies. The rather small sample size and observational nature of the study are further limitations.

WIDER IMPLICATIONS OF THE FINDINGS

Our results suggest the importance of pre-pregnancy weight loss in overweight women with PCOS in order to reduce the risk of adverse perinatal outcomes.

STUDY FUNDING/COMPETING INTERESTS

Veerle De Frène is holder of a Special PhD Fellowship by the Flemish Foundation for Scientific Research (FWO-Vlaanderen). Petra De Sutter is holder of a fundamental clinical research mandate by the Flemish Foundation for Scientific Research (FWO-Vlaanderen). There are no competing interests.

On regression adjustment for the propensity score

Propensity scores are widely adopted in observational research because they enable adjustment for high-dimensional confounders without requiring models for their association with the outcome of interest. The results of statistical analyses based on stratification, matching or inverse weighting by the propensity score are therefore less susceptible to model extrapolation than those based solely on outcome regression models. This is attractive because extrapolation in outcome regression models may be alarming, yet difficult to diagnose, when the exposed and unexposed individuals have very different covariate distributions. Standard regression adjustment for the propensity score forms an alternative to the aforementioned propensity score methods, but the benefits of this are less clear because it still involves modelling the outcome in addition to the propensity score. In this article, we develop novel insights into the properties of this adjustment method. We demonstrate that standard tests of the null hypothesis of no exposure effect (based on robust variance estimators), as well as particular standardised effects obtained from such adjusted regression models, are robust against misspecification of the outcome model when a propensity score model is correctly specified; they are thus not vulnerable to the aforementioned problem of extrapolation. We moreover propose efficient estimators for these standardised effects, which retain a useful causal interpretation even when the propensity score model is misspecified, provided the outcome regression model is correctly specified.

Mediation Analysis with Multiple Mediators

Recent advances in the causal inference literature on mediation have extended traditional approaches to direct and indirect effects to settings that allow for interactions and non-linearities. In this paper, these approaches from causal inference are further extended to settings in which multiple mediators may be of interest. Two analytic approaches, one based on regression and one based on weighting are proposed to estimate the effect mediated through multiple mediators and the effects through other pathways. The approaches proposed here accommodate exposure-mediator interactions and, to a certain extent, mediator-mediator interactions as well. The methods handle binary or continuous mediators and binary, continuous or count outcomes. When the mediators affect one another, the strategy of trying to assess direct and indirect effects one mediator at a time will in general fail; the approach given in this paper can still be used. A characterization is moreover given as to when the sum of the mediated effects for multiple mediators considered separately will be equal to the mediated effect of all of the mediators considered jointly. The approach proposed in this paper is robust to unmeasured common causes of two or more mediators.

On adjustment for auxiliary covariates in additive hazard models for the analysis of randomized experiments

We consider additive hazard models (Aalen, 1989) for the effect of a randomized treatment on a survival outcome, adjusting for auxiliary baseline covariates. We demonstrate that the Aalen least squares estimator of the treatment effect parameter is asymptotically unbiased, even when the hazard's dependence on time or on the auxiliary covariates is misspecified, and even away from the null hypothesis of no treatment effect. We moreover show that adjustment for auxiliary baseline covariates does not change the asymptotic variance of the Aalen least squares estimator of the effect of a randomized treatment. We conclude that, in view of its robustness against model misspecification, Aalen least squares estimation is attractive for evaluating treatment effects on a survival outcome in randomized experiments, and that the primary reasons to consider baseline covariate adjustment in such settings may be the interest in subgroup effects, or the need to adjust for informative censoring or for baseline imbalances. Our results also shed light on the robustness of Aalen least squares estimators against model misspecification in observational studies.

Predictability of cerebral palsy in a high-risk NICU population

AIM

This study aims to create a predictive model for the assessment of the individual risk of developing cerebral palsy in a large cohort of selected high-risk infants.

PATIENTS AND METHODS

1099 NICU-admitted high-risk infants were assessed up to the corrected age of at least 12 months. CP was categorized relative to subtype, distribution and severity. Several perinatal characteristics (gender, gestational age, multiple gestation, small for gestational age, perinatal asphyxia and duration of mechanical ventilation), besides neonatal cerebral ultrasound data were used in the logistic regression model for the risk of CP.

RESULTS

Perinatal asphyxia, mechanical ventilation>7 days, white matter disease except for transient echodensities<7 days, intraventricular haemorrhage grades III and IV, cerebral infarction and deep grey matter lesions were recognized as independent predictors for the development of CP. 95% of all children with CP were correctly identified at or above the cut-off value of 4.5% probability of CP development. Higher gestational age, perinatal asphyxia and deep grey matter lesion are independent predictors for non-spastic versus spastic CP (OR=1.1, 3.6, and 7.5, respectively). Independent risk factors for prediction of unilateral versus bilateral spastic CP are higher gestational age, cerebral infarction and parenchymal haemorrhagic infarction (OR=1.2, 31, and 17.6, respectively). Perinatal asphyxia is the only significant variable retained for the prediction of severe CP versus mild or moderate CP.

CONCLUSION

The presented model based on perinatal characteristics and neonatal US-detected brain injuries is a useful tool in identifying specific infants at risk for developing CP.

A randomized, double-blind, placebo-controlled, cross-over pilot study on the use of a standardized hop extract to alleviate menopausal discomforts

OBJECTIVES

To examine the efficacy of a hop extract (standardized at 100mug 8-prenylnaringenin per day) for relief of menopausal discomforts.

METHODS

A 16-week randomized, double-blind, placebo-controlled, cross-over study was conducted with 36 menopausal women. The participants were randomly allocated to either placebo or active treatment (hop extract) for a period of eight weeks after which treatments were switched for another eight weeks. The Kupperman Index (KI), the Menopause Rating Scale (MRS) and a multifactorial Visual Analogue Scale (VAS) were assessed at baseline, and after eight and sixteen weeks.

RESULTS

After 8 weeks, both active treatment and placebo significantly improved all outcome measures when compared to baseline with somewhat higher average reductions for placebo than for the active treatment. After 16 weeks only the active treatment after placebo further reduced all outcome measures, whereas placebo after active treatment resulted in an increase for all outcome measures. Although, the overall estimates of treatment efficacy (active treatment-placebo) based on linear mixed models do not show a significant effect, time-specific estimates of treatment efficacy indicate significant reductions for KI (P = 0.02) and VAS (P = 0.03) and a marginally significant reduction (P = 0.06) for MRS after 16 weeks.

CONCLUSIONS

Whereas the first treatment period resulted in similar reductions in menopausal discomforts in both treatment groups, results from the second treatment period suggest superiority of the standardized hop extract over placebo. Thus, phytoestrogen preparations containing this standardized hop extract may provide an interesting alternative to women seeking relief of mild vasomotor symptoms.

Support vector machine versus logistic regression modeling for prediction of hospital mortality in critically ill patients with haematological malignancies

Background

Several models for mortality prediction have been constructed for critically ill patients with haematological malignancies in recent years. These models have proven to be equally or more accurate in predicting hospital mortality in patients with haematological malignancies than ICU severity of illness scores such as the APACHE II or SAPS II [1]. The objective of this study is to compare the accuracy of predicting hospital mortality in patients with haematological malignancies admitted to the ICU between models based on multiple logistic regression (MLR) and support vector machine (SVM) based models.

Methods

352 patients with haematological malignancies admitted to the ICU between 1997 and 2006 for a life-threatening complication were included. 252 patient records were used for training of the models and 100 were used for validation. In a first model 12 input variables were included for comparison between MLR and SVM. In a second more complex model 17 input variables were used. MLR and SVM analysis were performed independently from each other. Discrimination was evaluated using the area under the receiver operating characteristic (ROC) curves (± SE).

Results

The area under ROC curve for the MLR and SVM in the validation data set were 0.768 (± 0.04) vs. 0.802 (± 0.04) in the first model (p = 0.19) and 0.781 (± 0.05) vs. 0.808 (± 0.04) in the second more complex model (p = 0.44). SVM needed only 4 variables to make its prediction in both models, whereas MLR needed 7 and 8 variables in the first and second model respectively.

Conclusion

The discriminative power of both the MLR and SVM models was good. No statistically significant differences were found in discriminative power between MLR and SVM for prediction of hospital mortality in critically ill patients with haematological malignancies.

The decline of serum testosterone levels in community-dwelling men over 70 years of age: descriptive data and predictors of longitudinal changes

OBJECTIVE

This study was designed to assess longitudinal changes in serum testosterone levels, explore relationships with aging, genetic-, health-, and lifestyle-related factors, and investigate predictors of changes in healthy elderly men.

DESIGN

Population-based, longitudinal, 4-year observational study in 221 community-dwelling men aged 71-86 years at baseline.

METHODS

Hormone levels assessed by immunoassay, anthropometry, questionnaires on general health, and genetic polymorphisms. Predictors of changes in testosterone levels explored using linear mixed-effects modeling for longitudinal analyses.

RESULTS

Total testosterone (TT), free testosterone, and bioavailable testosterone (BioT) levels decreased with aging, decreases in BioT being most marked. No changes in sex hormone-binding globulin (SHBG) or estradiol (E(2)), while LH and FSH levels increased during follow-up. Subjects who gained weight displayed a greater decline in TT levels, mainly due to decreasing SHBG levels. However, baseline body composition was not predictive of subsequent changes in testosterone levels. Baseline E(2) (P=0.023 to 0.004), LH (P=0.046 to 0.005), and FSH (P<0.002) levels were independently positively associated with a faster decline in testosterone fractions, although only FSH remained significant when adjusting for baseline testosterone (P=0.041-0.035). Carriers of a 'TA' haplotype of the estrogen receptor alpha gene (ER alpha) PvuII and XbaI polymorphisms displayed a slower decline of TT and BioT (P=0.041-0.007).

CONCLUSIONS

In elderly men with already low serum testosterone levels, a further decline was observed, independent of baseline age. The identification of FSH levels as a predictor of this decline appears to reflect the testicular mechanisms of aging-related changes in testosterone production, whereas associations with E(2) and ER alpha polymorphisms are suggestive of estrogen-related processes, possibly related to changes in the neuroendocrine regulation of testosterone production.

Natural history and clearance of HPV after treatment of precancerous cervical lesions

AIM

To assess the clearance rate of human papillomavirus (HPV) after out-patient treatment of cervical intraepithelial neoplasia (CIN).

METHODS AND RESULTS

A total of 122 Nicaraguan women with HPV DNA-positive and histologically confirmed CIN lesions were included in the study. Fifty-five patients with CIN1 and 67 with CIN2-3 were treated by cryotherapy and loop electrosurgical excision procedure (LEEP), respectively. Follow-up visits were scheduled at 6 weeks, 6 months, 1 year and 2 years. Investigations included cytology, HPV DNA testing and colposcopy/biopsy if needed. The clearance rate of HPV was calculated by multivariate logistic regression. Immediately after treatment, a pronounced decrease in presence of HPV was observed in both groups, with a significantly higher clearance in the LEEP group than in the cryotherapy group (P = 0.019). Subsequently, clearance continued over time and was similar between the cryotherapy group and the LEEP group (P = 0.73). Approximately the same detection rates were obtained for persistence of all HPV types and for high-risk types separately: 43.9, 37.6, 29.9 and 17.7% in the cryotherapy group and 24.9, 20.3, 15.3 and 8.4% in the LEEP group at 6 weeks, 6 months, 1 year and 2 years, respectively.

CONCLUSIONS

Out-patient treatment of precancerous lesions of the cervix usually results in clearance of HPV. Both LEEP and cryotherapy are highly effective methods of eradicating HPV. HPV DNA testing may have added value in the follow-up of patients.

Sense and sensitivity when correcting for observed exposures in randomized clinical trials

Standard intent-to-treat analyses of randomized clinical trials can yield biased estimates of treatment efficacy and toxicity when not all patients comply with their assigned treatment. Flexible methods have been proposed which correct for this by modelling expected contrasts between an individual's observed outcome and his/her potential outcome in the absence of exposure. Because such comparisons often require untestable assumptions, a sensitivity analysis is warranted. We show how this can be performed in a meaningful and practically useful way. Following the approach of Molenberghs, Kenward and Goetghebeur in a missing data context, we evaluate the separate contributions of structural uninformativeness and sampling variation to uncertainty about the population parameters. This leads us to consider Honestly Estimated Ignorance Regions (HEIRs) and Estimated Uncertainty RegiOns (EUROs), respectively. We use the results to estimate the causal effect of observed exposure on successful blood pressure reduction in a randomized controlled clinical trial with partial non-compliance.

Accounting for correlation and compliance in cluster randomized trials

This paper discusses causal inference with survival data from cluster randomized trials. It is argued that cluster randomization carries the potential for post-randomization exposures which involve differentially selective compliance between treatment arms, even for an all or nothing exposure at the individual level. Structural models can be employed to account for post-randomization exposures, but should not ignore clustering. We show how marginal modelling and random effects models allow to adapt structural estimators to account for clustering. Our findings are illustrated with data from a vitamin A trial for the prevention of infant mortality in the rural plains of Nepal.

Regression models for disease prevalence with diagnostic tests on pools of serum samples

Whether the aim is to diagnose individuals or estimate prevalence, many epidemiological studies have demonstrated the successful use of tests on pooled sera. These tests detect whether at least one sample in the pool is positive. Although originally designed to reduce diagnostic costs, testing pools also lowers false positive and negative rates in low prevalence settings and yields more precise prevalence estimates. Current methods are aimed at estimating the average population risk from diagnostic tests on pools. In this article, we extend the original class of risk estimators to adjust for covariates recorded on individual pool members. Maximum likelihood theory provides a flexible estimation method that handles different covariate values in the pool, different pool sizes, and errors in test results. In special cases, software for generalized linear models can be used. Pool design has a strong impact on precision and cost efficiency, with covariate-homogeneous pools carrying the largest amount of information. We perform joint pool and sample size calculations using information from individual contributors to the pool and show that a good design can severely reduce cost and yet increase precision. The methods are illustrated using data from a Kenyan surveillance study of HIV. Compared to individual testing, age-homogeneous, optimal-sized pools of average size seven reduce cost to 44% of the original price with virtually no loss in precision.