Search

Motivation Identifying cis-acting genetic variants associated with gene expression levels—an analysis commonly referred to as expression quantitative trait loci (eQTLs) mapping—is an important first step toward understanding the genetic determinant of gene expression variation. Successful eQTL mapping requires effective control of confounding factors. A common method for confounding effects control in eQTL mapping studies is the probabilistic estimation of expression residual (PEER) analysis. PEER analysis extracts PEER factors to serve as surrogates for confounding factors, which is further included in the subsequent eQTL mapping analysis. However, it is computationally challenging to determine the optimal number of PEER factors used for eQTL mapping. In particular, the standard approach to determine the optimal number of PEER factors examines one number at a time and chooses a number that optimizes eQTLs discovery. Unfortunately, this standard approach involves multiple repetitive eQTL mapping procedures that are computationally expensive, restricting its use in large-scale eQTL mapping studies that being collected today. Results Here, we present a simple and computationally scalable alternative, Effect size Correlation for COnfounding determination (ECCO), to determine the optimal number of PEER factors used for eQTL mapping studies. Instead of performing repetitive eQTL mapping, ECCO jointly applies differential expression analysis and Mendelian randomization analysis, leading to substantial computational savings. In simulations and real data applications, we show that ECCO identifies a similar number of PEER factors required for eQTL mapping analysis as the standard approach but is two orders of magnitude faster. The computational scalability of ECCO allows for optimized eQTL discovery across 48 GTEx tissues for the first time, yielding an overall 5.89% power gain on the number of eQTL harboring genes (eGenes) discovered as compared to the previous GTEx recommendation that does not attempt to determine tissue-specific optimal number of PEER factors. Availabilityand implementation Our method is implemented in the ECCO software, which, along with its GTEx mapping results, is freely available at www.xzlab.org/software.html. All R scripts used in this study are also available at this site. Supplementary information Supplementary data are available at Bioinformatics online.

BackgroundVision loss may be a risk factor for hallucinations, but this has not been studied at the population level.MethodsTo determine the association between self-reported vision loss and hallucinations in a large community-based sample of older adults, we performed a cross-sectional and longitudinal analysis of two large, nationally representative US health surveys: the National Health and Aging Trends Study (NHATS) and the Health and Retirement Study (HRS). Visual impairment and hallucinations were self- or proxy-reported. Multivariate single and mixed effects logistic regression models were built to examine whether visual impairment and history of cataract surgery were associated with hallucinations.ResultsIn NHATS (n = 1520), hallucinations were more prevalent in those who reported difficulty reading newspaper print (OR 1.77, 95% confidence interval (CI): 1.32–2.39) or recognising someone across the street (OR 2.48, 95% CI: 1.86–3.31) after adjusting for confounders. In HRS (n = 3682), a similar association was observed for overall (OR 1.32, 95% CI: 1.08–1.60), distance (OR 1.61, 95% CI: 1.32–1.96) and near eyesight difficulties (OR 1.52, 95% CI: 1.25–1.85). In neither sample was there a significant association between cataract surgery and hallucinations after adjusting for covariates.ConclusionsVisual dysfunction is associated with increased odds of hallucinations in the older US adult population. This suggests that the prevention and treatment of vision loss may potentially reduce the prevalence of hallucinations in older adults.

BACKGROUND: increasing numbers of older adults are living with frailty and its adverse consequences. We investigated relationships between diet quality or patterns and incident physical frailty in older British men and whether any associations were influenced by inflammation. METHODS: prospective study of 945 men from the British Regional Heart Study aged 70–92 years with no prevalent frailty. Incident frailty was assessed by questionnaire after 3 years of follow-up. Frailty was defined as having at least three of: low grip strength, low physical activity, slow walking speed, unintentional weight loss and feeling of low energy, all based on self-report. The Healthy Diet Indicator (HDI) based on WHO dietary guidelines and the Elderly Dietary Index (EDI) based on a Mediterranean-style dietary intake were computed from questionnaire data and three dietary patterns were identified using principal components analysis: prudent, high fat/low fibre and high sugar. RESULTS: men in the highest EDI category and those who followed a prudent diet were less likely to become frail [top vs bottom category odds ratio (OR) (95% CI) 0.49 (0.30, 0.82) and 0.53 (0.30, 0.92) respectively] after adjustment for potential confounders including BMI and prevalent cardiovascular disease. No significant association was seen for the HDI. By contrast those who had a high fat low fibre diet pattern were more likely to become frail [OR (95% CI) 2.54 (1.46, 4.40)]. These associations were not mediated by C-reactive protein (marker of inflammation). CONCLUSIONS: the findings suggest adherence to a Mediterranean–style diet is associated with reduced risk of developing frailty in older people.

Motivation Differential coexpression—the alteration of gene coexpression patterns observed in different biological conditions—has been proposed to be a mechanism for revealing rewiring of transcription regulatory networks. Despite wide use of methods for differential coexpression analysis, the phenomenon has not been well-studied. In particular, in many applications, differential coexpression is confounded with differential expression, that is, changes in average levels of expression across conditions. This confounding, despite affecting the interpretation of the differential coexpression, has rarely been studied. Results We constructed high-quality coexpression networks for five human tissues and identified coexpression links (gene pairs) that were specific to each tissue. Between 3 and 32% of coexpression links were tissue-specific (differentially coexpressed) and this specificity is reproducible in an external dataset. However, we show that up to 75% of the observed differential coexpression is substantially explained by average expression levels of the genes. ‘Pure’ differential coexpression independent from differential expression is a minority and is less reproducible in external datasets. We also investigated the functional relevance of pure differential coexpression. Our conclusion is that to a large extent, differential coexpression is more parsimoniously explained by changes in average expression levels and pure links have little impact on network-based functional analysis. Supplementary information Supplementary data are available at Bioinformatics online.

Motivation For many traits, causal loci uncovered by genetic mapping studies explain only a minority of the heritable contribution to trait variation. Multiple explanations for this ‘missing heritability’ have been proposed. Single nucleotide polymorphism (SNP)–SNP interaction (epistasis), as one of the compelling models, has been widely studied. However, the genome-wide scan of epistasis, especially for quantitative traits, poses huge computational challenges. Moreover, covariate adjustment is largely ignored in epistasis analysis due to the massive extra computational undertaking. Results In the current study, we found striking differences among epistasis models using both simulation data and real biological data, suggesting that not only can covariate adjustment remove confounding bias, it can also improve power. Furthermore, we derived mathematical formulas, which enable the exhaustive epistasis scan together with full covariate adjustment to be expressed in terms of large matrix operation, therefore substantially improving the computational efficiency (∼104× faster than existing methods). We call the new method MatrixEpistasis. With MatrixEpistasis, we re-analyze a large real yeast dataset comprising 11 623 SNPs, 1008 segregants and 46 quantitative traits with covariates fully adjusted and detect thousands of novel putative epistasis with P-values Availability and implementation The method is implemented in R and available at https://github.com/fanglab/MatrixEpistasis. Supplementary information Supplementary data are available at Bioinformatics online.

Motivation: Repeated cross-sectional time series single cell data confound several sources of variation, with contributions from measurement noise, stochastic cell-to-cell variation and cell progression at different rates. Time series from single cell assays are particularly susceptible to confounding as the measurements are not averaged over populations of cells. When several genes are assayed in parallel these effects can be estimated and corrected for under certain smoothness assumptions on cell progression. Results: We present a principled probabilistic model with a Bayesian inference scheme to analyse such data. We demonstrate our method’s utility on public microarray, nCounter and RNA-seq datasets from three organisms. Our method almost perfectly recovers withheld capture times in an Arabidopsis dataset, it accurately estimates cell cycle peak times in a human prostate cancer cell line and it correctly identifies two precocious cells in a study of paracrine signalling in mouse dendritic cells. Furthermore, our method compares favourably with Monocle, a state-of-the-art technique. We also show using held-out data that uncertainty in the temporal dimension is a common confounder and should be accounted for in analyses of repeated cross-sectional time series. Availability and Implementation: Our method is available on CRAN in the DeLorean package. Contact: john.reid@mrc-bsu.cam.ac.uk Supplementary information: Supplementary data are available at Bioinformatics online.

Background: animal studies suggest a neuroprotective role for leptin, but human studies have shown mixed results. We examined whether plasma leptin levels in individuals with mild cognitive impairment (MCI) were related to cognitive function at baseline and whether higher leptin levels were associated with reduced risk of dementia. Methods: we categorised 352 MCI participants into sex-specific tertiles based on log-transformed fasting plasma leptin levels. In sex-stratified analyses, we investigated whether cognitive ability differed by leptin tertile. We also examined whether the risk of dementia over a 3-year follow-up period differed by leptin level. Analyses controlled for numerous potential confounding variables, including body mass index, hypertension and levels of blood insulin and C-reactive protein. Results: baseline cognitive ability did not differ as a function of leptin level, nor were higher leptin levels associated with reduced hazard of developing dementia. Controlling for related co-variates did not reveal any significant associations between leptin and dementia risk. Conclusion: in this cohort of older adults with MCI, plasma leptin level was not associated with cognitive function at baseline, nor did it predict risk of dementia. Other biological measures, such as volumetric MRI and cerebrospinal fluid protein levels, have demonstrated robust dementia prediction in this cohort. Thus, the current negative findings suggest that plasma leptin, on its own, is unlikely to become a useful clinical biomarker for Alzheimer's disease. Efforts to develop other blood-based biomarkers are needed.

Background fear of falling (FOF) is a major health concern among community-dwelling older adults that could restrict mobility. Objective to examine the association of FOF with life-space mobility (i.e. the spatial area a person moves through in daily life) of community-dwelling older adults from five diverse sites. Methods in total, 1,841 older adults (65-74 years) were recruited from Kingston, Canada; Saint-Hyacinthe, Canada; Tirana, Albania; Manizales, Colombia and Natal, Brazil. FOF was assessed using the Fall Efficacy Scale-International (FES-I total score), and the life space was quantified using the Life-Space Assessment (LSA), a scale that runs from 0 (minimum life space) to 120 (maximum life space). Results the overall average LSA total score was 68.7 (SD: 21.2). Multiple-linear regression analysis demonstrated a significant relationship of FOF with life-space mobility, even after adjusting for functional, clinical and sociodemographic confounders (B = -0.15, 95% confidence interval (CI) -0.26 to -0.04). The FOF × site interaction term was significant with a stronger linear relationship found in the Canadian sites and Tirana compared with the South American sites. After adjusting for all confounders, the association between FOF with LSA remained significant at Kingston (B = -0.32, 95% CI -0.62 to -0.01), Saint-Hyacinthe (B = -0.81, 95% CI -1.31 to -0.32) and Tirana (B = -0.57, 95% CI -0.89 to -0.24). Conclusion FOF is an important psychological factor that is associated with reduction in life space of older adults in different social and cultural contexts, and the strength of this association is site specific. Addressing FOF among older adults would help improve their mobility in local communities, which in turn would improve social participation and health-related quality of life.

High-throughput studies have been extensively conducted in the research of complex human diseases. As a representative example, consider gene-expression studies where thousands of genes are profiled at the same time. An important objective of such studies is to rank the diagnostic accuracy of biomarkers (e.g. gene expressions) for predicting outcome variables while properly adjusting for confounding effects from low-dimensional clinical risk factors and environmental exposures. Existing approaches are often fully based on parametric or semi-parametric models and target evaluating estimation significance as opposed to diagnostic accuracy. Receiver operating characteristic (ROC) approaches can be employed to tackle this problem. However, existing ROC ranking methods focus on biomarkers only and ignore effects of confounders. In this article, we propose a model-based approach which ranks the diagnostic accuracy of biomarkers using ROC measures with a proper adjustment of confounding effects. To this end, three different methods for constructing the underlying regression models are investigated. Simulation study shows that the proposed methods can accurately identify biomarkers with additional diagnostic power beyond confounders. Analysis of two cancer gene-expression studies demonstrates that adjusting for confounders can lead to substantially different rankings of genes.

Motivation: Adjustment for population structure is necessary to avoid bias in genetic association studies of susceptibility variants for complex diseases. Population structure may differ from one genomic region to another due to the variability of individual ancestry associated with migration, random genetic drift or natural selection. Current association methods for correcting population stratification usually involve adjustment of global ancestry between study subjects. Results: We suggest interrogating local population structure for fine mapping to more accurately locate true casual genes by better adjusting the confounding effect due to local ancestry. By extensive simulations on genome-wide datasets, we show that adjusting global ancestry may lead to false positives when local population structure is an important confounding factor. In contrast, adjusting local ancestry can effectively prevent false positives due to local population structure and thus can improve fine mapping for disease gene localization. We applied the local and global adjustments to the analysis of datasets from three genome-wide association studies, including European Americans, African Americans and Nigerians. Both European Americans and African Americans demonstrate greater variability in local ancestry than Nigerians. Adjusting local ancestry successfully eliminated the known spurious association between SNPs in the LCT gene and height due to the population structure existed in European Americans. Contact: xiaofeng.zhu@case.edu Supplementary information: Supplementary data are available at Bioinformatics online.

Background: Measuring quality of care and ranking hospitals with outcome measures poses two major methodological challenges: case-mix adjustment and variation that exists by chance. Aim: To compare methods for comparing and ranking hospitals that considers these. Methods: The Netherlands Stroke Survey was conducted in 10 hospitals in the Netherlands, between October 2002 and May 2003, with prospective and consecutive enrolment of patients with acute brain ischaemia. Poor outcome was defined as death or disability after 1 year (modified Rankin scale of ⩾3). We calculated fixed and random hospital effects on poor outcome, unadjusted and adjusted for patient characteristics. We compared the hospitals using the expected rank, a novel statistical measure incorporating the magnitude and the uncertainty of differences in outcome. Results: At 1 year after stroke, 268 of the total 505 patients (53%) had a poor outcome. There were substantial differences in outcome between hospitals in unadjusted analysis (χ2 = 48, 9 d f , P < 0.0001). Adjustment for 12 confounders led to halving of the χ2 (χ2 = 24). The same pattern was observed in random effects analysis. Estimated performance of individual hospitals changed considerably between unadjusted and adjusted analysis. Further changes were seen with random effect estimation, especially for smaller hospitals. Ordering by expected rank led to shrinkage of the original ranks of 1–10 towards the median rank of 5.5 and to a different order of the hospitals, compared to ranking based on fixed effects. Conclusion: In comparing and ranking hospitals, case-mix-adjusted random effect estimates and the expected ranks are more robust alternatives to traditional fixed effect estimates and simple rankings.

Background: the identification of modifiable risk factors for preventing disability in older individuals is essential for planning preventive strategies. Purpose: to identify cross-sectional correlates of disability and risk factors for the development activities of daily living (ADL) and instrumental ADL (IADL) disability in community-dwelling older adults. Methods: the study population consisted of 897 subjects aged 65–102 years from the InCHIANTI study, a population-based cohort in Tuscany (Italy). Factors potentially associated with high risk of disability were measured at baseline (1998–2000), and disability in ADLs and IADLs were assessed both at baseline and at the 3-year follow-up (2001–03). Results: the baseline prevalence of ADL disability and IADL disability were, respectively, 5.5% (49/897) and 22.2% (199/ 897). Of 848 participants free of ADL disability at baseline, 72 developed ADL disability and 25 of the 49 who were already disabled had a worsening in ADL disability over a 3-year follow-up. Of 698 participants without IADL disability at baseline, 100 developed IADL disability and 104 of the 199 who already had IADL disability had a worsening disability in IADL over 3 years. In a fully adjusted model, high level of physical activity compared to sedentary state was significantly associated with lower incidence rates of both ADL and IADL disability at the 3-year follow-up visit (odds ratio (OR): 0.30; 95% confidence intervals (CI) 0.12–0.76 for ADL disability and OR: 0.18; 95% CI 0.09–0.36 for IADL disability). After adjusting for multiple confounders, higher energy intake (OR for difference in 100kcal/day: 1.09; 95% CI 1.02–1.15) and hypertension (OR: 1.91; 95% CI 1.06–3.43) were significant risk factors for incident or worsening ADL disability. Conclusions: higher level of physical activity and lower energy intake may be protective against the development in ADL and IADL disability in older persons.

Background: previous studies examining the relationship between socioeconomic status and telomere length showed conflicting results, one study finding shorter telomere length in subjects with lower socioeconomic status and one showing no relationship. Design: cross-sectional study. Setting: community-living elderly Chinese in Hong Kong. Objective: this study examines the relationship between self-rated social economic status and telomere length in Hong Kong Chinese men and women aged 65 years and over living in the community. Subjects and method: information was collected from 958 men and 978 women regarding possible confounding factors such as the presence of chronic diseases, smoking, physical activity level, dietary intake and body mass index. Telomere length was measured by quantitative PCR. Result: in men only, after adjustment for age and other confounding factors, a higher ranking in community standing was associated with shorter telomere length. Conclusion: men with higher self-rated socioeconomic status have shorter telomeres, possibly mediated through psychosocial rather than lifestyle factors or the presence of chronic disease. There may be cultural ethnic and age-related differences in social determinants of health.

Aims To assess long-term changes in left atrial (LA) volume in patients with morbid obesity [body mass index (BMI) ≥35 kg/m2 with co-morbidities] and extreme obesity (BMI ≥40 kg/m2), after surgically-induced weight loss (WL) after gastric bypass surgery. Methods and results We reviewed 57 patients who underwent gastric bypass surgery and had echocardiograms both before and after the operation. A control group was frequency-matched for BMI, sex, age, and for duration of follow-up. After a mean follow-up of 3.6 years, LA volume did not change significantly in patients who underwent bariatric surgery, but increased in the control group by 15 ± 28 ml ( P < 0.0001), and 0.1 ± 0.2 ml ( P < 0.0001) for height-indexed LA volume, with a difference between cases and controls that remained significant after adjusting for potential confounders ( P = 0.01). In the study population as a whole, there was a positive correlation between change in body weight and change in LA volume ( r = 0.22, P = 0.006) independent of clinical conditions associated with LA enlargement. Conclusion Change in body weight is associated with change in LA size independent of obesity-associated co-morbidities. Successful WL induced by bariatric surgery prevents the progressive increase in LA volume.

Background Quantitative bias analysis (QBA) measures study errors in terms of direction, magnitude and uncertainty. This systematic review aimed to describe how QBA has been applied in epidemiological research in 2006–19. Methods We searched PubMed for English peer-reviewed studies applying QBA to real-data applications. We also included studies citing selected sources or which were identified in a previous QBA review in pharmacoepidemiology. For each study, we extracted the rationale, methodology, bias-adjusted results and interpretation and assessed factors associated with reproducibility. Results Of the 238 studies, the majority were embedded within papers whose main inferences were drawn from conventional approaches as secondary (sensitivity) analyses to quantity-specific biases (52%) or to assess the extent of bias required to shift the point estimate to the null (25%); 10% were standalone papers. The most common approach was probabilistic (57%). Misclassification was modelled in 57%, uncontrolled confounder(s) in 40% and selection bias in 17%. Most did not consider multiple biases or correlations between errors. When specified, bias parameters came from the literature (48%) more often than internal validation studies (29%). The majority (60%) of analyses resulted in >10% change from the conventional point estimate; however, most investigators (63%) did not alter their original interpretation. Degree of reproducibility related to inclusion of code, formulas, sensitivity analyses and supplementary materials, as well as the QBA rationale. Conclusions QBA applications were rare though increased over time. Future investigators should reference good practices and include details to promote transparency and to serve as a reference for other researchers.

In their seminal 2002 paper, “Causal Knowledge as a Prerequisite for Confounding Evaluation: An Application to Birth Defects Epidemiology,” Hernán et al. emphasized the importance of using theory rather than data to guide confounding control, focusing on colliders as variables that share characteristics with confounders but whose control may actually introduce bias into analyses. In this commentary, we propose that the importance of this paper stems from the connection that the authors made between non-exchangeability as the ultimate source of bias and structural representations of bias using directed acyclic graphs. This provided both a unified approach to conceptualizing bias and a means of distinguishing between different sources of bias, in particular confounding and selection bias. Drawing on examples from the paper, we also highlight unresolved questions about the relationship between collider bias, selection bias, and generalizability and argue that causal knowledge is not only a prerequisite for identifying confounders but also for developing any hypothesis about potential sources of bias.

BACKGROUNDPathology laboratories are required to immediately report results which indicate a patient is at critical risk, but there is little consensus about what values are deemed critical. The aim of this review was to systematically review the literature on alert thresholds for common chemistry and hematology tests in adults and to provide an explicit and ranked source of this evidence.METHODSThe literature search covered the period of 1995–2014. Evidence sources were critically appraised and ranked using the 1999 Stockholm hierarchy for analytical performance specifications in laboratory medicine modified for establishing decision limits.RESULTSThe 30 most frequently reported laboratory tests with alert thresholds are presented with evidence rankings. Similar thresholds were reported in North America, Europe and Asia. Seventy percent of papers reported thresholds set by individual institutions, while 18% contained thresholds from surveys of laboratories or clinicians. Forty-six percent of the papers referred to 1 or both of the 2 American laboratory surveys from the early 1990s. “Starter sets” of alert thresholds were recommended by 6 professional bodies, 3 of which were collaborations between pathologists and clinicians. None of the 9 outcome studies identified dealt with confounding factors.CONCLUSIONSRecommendations by professional bodies based on outdated surveys of the former state of the art or consensus are currently the best sources of evidence for laboratories to build their alert list. Well-designed outcome studies and greater collaboration between clinicians and the laboratory are needed to identify the most appropriate alert thresholds that signify actionable, critical or significant risk to patient well-being.

In a 1993 paper (Am J Epidemiol. 1993;137(1):1–8), Weinberg considered whether a variable that is associated with the outcome and is affected by exposure but is not an intermediate variable between exposure and outcome should be considered a confounder in etiologic studies. As an example, she examined the common practice of adjusting for history of spontaneous abortion when estimating the effect of an exposure on the risk of spontaneous abortion. She showed algebraically that such an adjustment could substantially bias the results even though history of spontaneous abortion would meet some definitions of a confounder. Directed acyclic graphs (DAGs) were introduced into epidemiology several years later as a tool with which to identify confounders. The authors now revisit Weinberg's paper using DAGs to represent scenarios that arise from her original assumptions. DAG theory is consistent with Weinberg's finding that adjusting for history of spontaneous abortion introduces bias in her original scenario. In the authors' examples, treating history of spontaneous abortion as a confounder introduces bias if it is a descendant of the exposure and is associated with the outcome conditional on exposure or is a child of a collider on a relevant undirected path. Thoughtful DAG analyses require clear research questions but are easily modified for examining different causal assumptions that may affect confounder assessment.

Background: measures of physical capability may be predictive of subsequent health, but existing published studies have not been systematically reviewed. We hypothesised that weaker grip strength, slower walking speed and chair rising and shorter standing balance time, in community-dwelling populations, would be associated with higher subsequent risk of fracture, cognitive outcomes, cardiovascular disease, hospitalisation and institutionalisation. Methods: studies were identified through systematic searches of the electronic databases MEDLINE and EMBASE (to May 2009). Reference lists of eligible papers were also manually searched. Results: twenty-four papers had examined the associations between at least one physical capability measure and one of the outcomes. As the physical capability measures and outcomes had been assessed and categorised in different ways in different studies, and there were differences in the potential confounding factors taken into account, this made it impossible to pool results. There were more studies examining fractures than other outcomes, and grip strength and walking speed were the most commonly examined capability measures. Most studies found that weaker grip strength and slower walking speed were associated with increased risk of future fractures and cognitive decline, but residual confounding may explain results in some studies. Associations between physical capability levels and the other specified outcomes have not been tested widely. Conclusions: there is some evidence to suggest that objective measures of physical capability may be predictors of subsequent health in older community-dwelling populations. Most hypothesised associations have not been studied sufficiently to draw definitive conclusions suggesting the need for further research.

The power variance function distributions, which include the gamma and compound Poisson (CP) distributions among others, are commonly used in frailty models for family data. In a previous paper, we presented a frailty model constructed by randomizing the scale parameter in a CP distribution. When combined with a parametric baseline hazard, this yields a model with heterogeneity on both the individual and the family level and a subgroup with zero frailty, corresponding to people not experiencing the event. In this paper, we discuss covariates in the model. Depending on where the covariates are inserted in the model, one may have proportional hazards at the individual level, the family level, and a larger group level (for covariates shared by many families, e.g. ethnic groups) or get accelerated failure times. Each of these alternatives gives a specific interpretation of the covariate effects. An application to data infant mortality in siblings from the Medical Birth Registry of Norway is included. We compare the results for some of the different covariate modeling options.

Background: While there is good evidence to suggest an inverse relation of physical activity and cardiorespiratory fitness with coronary heart disease (CHD) in middle-aged men and women, much less is known about this association in older adults. The purpose of this paper was to explore the relation of physical activity and cardiorespiratory fitness with CHD in older adults by reviewing relevant studies. Methods: Publications were identified in two ways: i) conducting a PUBMED search from its inception in 1966 until January 2001; and ii) scrutinizing the reference sections of identified papers. Results: Ten studies relating physical activity and two relating cardiorespiratory fitness in older people to CHD met the inclusion criteria. With one exception, the studies were observational in nature and the majority of these were of prospective cohort design. Most studies featured men only. Of the eleven studies that presented data on older men, eight reported an inverse relation between physical activity or cardiorespiratory fitness and CHD, and statistical significance was seen in five of these. There were too few data on older women to draw clear conclusions regarding the association in this group. Conclusions: This review suggests that, in older adult men, physical activity and cardiorespiratory fitness are inversely related to CHD risk. This association is unlikely to be attributable to reverse causality or confounding. Except where such advice is contraindicated, older adult men may benefit from physical activity in terms of reduced CHD risk.

SummaryThe paper addresses a formal definition of a confounder based on the qualitative definition that is commonly used in standard epidemiology text-books. To derive the criterion for a factor to be a confounder given by Miettinen and Cook and to clarify inconsistency between various criteria for a confounder, we introduce the concepts of an irrelevant factor, an occasional confounder and a uniformly irrelevant factor. We discuss criteria for checking these and show that Miettinen and Cook's criterion can also be applied to occasional confounders. Moreover, we consider situations with multiple potential confounders, and we obtain two necessary conditions that are satisfied by each confounder set. None of the definitions and results presented in this paper require the ignorability and sufficient control confounding assumptions which are commonly employed in observational and epidemiological studies.

Ischaemic heart disease (IHD) is one of the more extensively researched diseases and most of its risk factors have been identified by consistent data from different studies. Diet and dietary patterns have been found to be correlated with the risk of IHD, for example, the dietary pattern of fruits and vegetables, whole-meal bread, low fat dairy and little alcohol has an inverse association with the risk, 1 while high intake of red and processed meat, refined grains and sugars, French fries and high fat dairy is associated with increased IHD risk. 2 Milk has been less consistently correlated with IHD and two recent reviews of published literature suggest there is no increased risk. 3,4 Milk is a rather complex food because it is a rich source of protein, saturated fat, calcium and lactose sugar. Saturated fat from dairy products, including whole milk has been found to be related to IHD. 5 Calcium from milk and dairy products is a major source of daily human consumption but the studies have failed to find an association with IHD. 6,7 Lactose is rather specific to unfermented milk and if found related to IHD could specifically point to unfermented milk as the causative nutrient. This is the basis of the paper by Siegel titled: Hypothesis: Is lactose a dietary risk factor for IHD? 8 published 28 years ago in this journal. The association between lactose and IHD was a reasonable and novel hypothesis to pursue three decades ago given the era of increasing awareness of IHD and the search for possible explanations from available data. This was similar to the study by Keys in 1980, 9 which suggested that saturated fat is associated with IHD based on correlational ecological data. However, a closer look at the paper by Segall reveals some methodological problems that undermine the eloquent argument for the observed association between lactose and IHD. This might also explain why no subsequent data was published to support this hypothesis. The paper does present the arguments for a hypothesis rather than a causal association, and therefore an ecological study design for the purpose of promoting such a hypothesis is acceptable. The term ‘ecological fallacy’ was coined in the 1950s and the bias introduced by attempting to apply findings from aggregate or geographical data to the individual-level is well-known. 10 Comparisons of countries without adjustment for important confounders between these countries, which are usually not available from aggregate data, can lead to false conclusions. The well known example of finding higher suicide rates in predominantly Protestant European countries compared with predominantly Catholic countries has been attributed to this fallacy. 11 In his paper, Segall does present the possibility that environmental factors explain the observed association given that the group of countries with high lactose exposure [low prevalence of lactose malabsorption also known as lactose intolerance (LI)] and high IHD were the developed countries with high socioeconomic status. This explanation is then dismissed based on low IHD in Japan and high IHD in the Punjab. The data on LI from the Punjab was based only on 18 subjects, and in the cited IHD study on Punjabis, 91% of the cases were from the high and middle income Punjabi social class, while the unique diet in Japan of fish and seafood high in omega-3 fatty acids (that is mostly protective against IHD) could explain why a country with high socioeconomic status would have low IHD. In a letter to the editor on the inverse association between wine and IHD in European countries, Segall argues about the ‘possible coincidental presence of a geographical factor of unknown identity that explains this association’, 12 an argument that can also explain the findings in the paper on milk and IHD published by Segall.

Dear Editor, The recent paper by Evangelista et al.1 seems to offer the promise of an inexpensive, readily available, easy to administer and highly efficacious adjunctive therapy for the treatment of chronic venous leg ulcers. We congratulate the authors on this very provocative study in which subjects were randomized to receive either simvastatin 40 mg daily or a placebo pill, in addition to standard compression therapy. As with any surprisingly positive finding, the potential benefit of added simvastatin will require independent confirmation, which should address some potentially important methodological issues. Firstly, this was a small single-centre trial. Regulatory agencies require multicentre trials for confirmation of efficacy in order to prevent unintentional or conscious bias, outlier outcomes or other idiosyncratic site effects from significantly influencing overall results. A single-centre study has none of these safeguards in place and therefore can only be considered exploratory. Secondly, it would be difficult to imagine more favourable outcomes than those achieved in this trial. The authors report healing rates of 67% at 10 weeks for ulcers > 20 cm2 (> 5 cm in diameter) and 100% for ulcers ≤ 20 cm2. This result is compared with the placebo group where 0% and 50% achieved closure, respectively. These results are particularly surprising when one considers that these ulcers were not small and not young; the mean size for ulcers that closed was 12 cm2 and 31 cm2 for placebo and simvastatin groups with average chronicity of 2·4 years and 3·3 years, respectively. According to Margolis et al.,2 a wound > 10 cm2 and > 12 months old has only a 22% chance of healing after 24 weeks of good therapy with compression. If one compares the outcomes achieved for the placebo group with those from other studies with a compression-only group, it is apparent that the proportion healed is unusually high. Kirsner et al.3 reported 46% healed at 12 weeks for ulcers treated with four-layer compression. Although this seems similar to the 50% reported for the smaller ulcers (≤ 20 cm2) in the Evangelista et al. study, these ulcers were all < 12 cm2 (mean = 5·6 cm2) and ≤ 2 years old. As Evangelista et al. point out, larger longer-duration ulcers are more difficult to heal. Why were these large, long-duration ulcers healed with such comparative facility? It is possible that this study represented the first application of good care (i.e. consistent, good-quality compression, infection control, control of oedema). The study did not employ a run-in period to exclude those who would heal under standard care alone. It is also not clear from their paper whether subjects were required to be either statin naive, or to have discontinued prestudy statin usage and undergone a ‘washout’. This is important because subjects with hypercholesterolaemia were allowed to enrol. While the number of affected subjects is not given, it would be reasonable to assume based on previous work4 that at least 12–20% had serum cholesterol of ≥ 240 mg dL−1 and would be categorized as having hypercholesterolaemia. If those subjects with ‘high-borderline’ serum cholesterol (200–239 mg dL−1) and ≥ 2 risk factors are included, it is likely to be closer to 30% of subjects.4 Presumably, most (if not all) subjects were under statin treatment prior to the study. Were subjects allowed to continue taking their prestudy statins or were they required to discontinue use? Allowing subjects to continue these medications would introduce several difficulties, including nonstandardized dosing in the active treatment group and a serious confounding variable for those randomized to the placebo group. The alternative, requiring cessation of statins before enrolling, would present challenges, particularly for those randomized to the placebo group. From a practical standpoint it would also put a serious limitation on the generalizability of these results as statins are in widespread common usage, particularly in this target population. Moreover, a subject undergoing chronic statin therapy who develops a venous ulcer would seem to be in some sense a ‘nonresponder’ to statin treatment, at least in regard to venous ulcers. From our own phase II trial comparing HP802-247 with compression alone,3 we know that of those subjects enrolling who were concomitantly undergoing statin therapy (n = 48, mean duration prestudy = 50·5 months), the overall healing rate at 12 weeks was only modestly higher than for those not on a statin (65% vs. 59%, not significant). If those subjects who were randomized to compression only and not treated with HP802-247 are considered, the difference is even smaller (50% vs. 45%, not significant). Several other important factors affecting healing were identified including ulcer area, chronicity and bacterial load.5 An inexpensive, highly efficacious, adjunctive treatment for venous leg ulcers, which could be administered as a once-daily tablet, is surely a very exciting prospect. Confirmation will require replicating these results in larger multicentre studies with appropriate inclusion and exclusion criteria, study design and complete transparency of these key trial design components.

Background Smoking rates in Greece are the highest recorded among OECD countries, but the historical and life-course evolution of smoking patterns is largely unknown. The present paper addresses this gap. Methods We produce nationally representative life-course trajectories of smoking and related mortality of eight generations of Greek men and women. We estimate the smoking–mortality correlation conditional on several confounders and project the estimates forward. Results We show that smoking prevalence among Greek men has plateaued at >60% for all but the youngest generation. For women, smoking prevalence is relatively lower, lags by several generations and follows a hump-shaped pattern. Smoking-attributable mortality is currently peaking for men (nearing 40% of total deaths) and is rising for women. We estimate that it takes ~20 years of smoking to maximize the smoking–mortality correlation (at 0.48 for men and 0.32 for women). Based on this estimation, we forecast that mortality rates will begin falling within the current decade. Conclusions The breadth of the Greek smoking epidemic has been high by international standards, reflecting the ineffective tobacco control efforts in the country. While smoking popularity fell during the Great Recession, policy vigilance is necessary to prevent a relapse once the economy recovers.

A synthesis of classical and recent thinking on the issues involved in selecting controls for case-control studies is presented in this and two companion papers (S. Wacholder et al. Am J Epidemiol 1992;135:1029-50). In this paper, a theoretical framework for selecting controls in case-control studies is developed. Three principles of comparability are described: 1) study base, that all comparisons be made within the study base; 2) deconfounding, that comparisons of the effects of the levels of exposure on disease risk not be distorted by the effects of other factors; and 3) comparable accuracy, that any errors in measurement of exposure be nondifferential between cases and controls. These principles, if adhered to in a study, can reduce selection, confounding, and information bias, respectively. The principles, however, are constrained by an additional efficiency principle regarding resources and time. Most problems and controversies in control selection reflect trade-offs among these four principles.

The test-negative study design is often used to estimate vaccine effectiveness in influenza studies, but it has also been proposed in the context of other infectious diseases, such as cholera, dengue, or Ebola. It was introduced as a variation of the case-control design, in an attempt to reduce confounding bias due to health-care–seeking behavior, and has quickly gained popularity because of its logistic advantages. However, examination of the directed acyclic graphs that describe the test-negative design reveals that without strong assumptions, the estimated odds ratio derived under this sampling mechanism is not collapsible over the selection variable, such that the results obtained for the sampled individuals cannot be generalized to the whole population. In this paper, we show that adjustment for severity of disease can reduce this bias and, under certain assumptions, makes it possible to unbiasedly estimate a causal odds ratio. We support our findings with extensive simulations and discuss them in the context of recently published cholera test-negative studies of the effectiveness of cholera vaccines.

We would like to thank Drs Ding, Hu and Pischon for their insights and commentaries on our paper ‘Dietary patterns and the risk of mortality: impact of cardiorespiratory fitness’. 1 In the spirit of a healthy debate, there are a few issues that we would like to address in our response. We would first like to highlight that the purpose of our paper was not to argue against the role of an unhealthy diet as a risk factor for morbidity and mortality. We are in full agreement that a healthy diet is important and that there is an abundance of evidence to support this position. Rather, our goal was to highlight that previous studies examining the relation between dietary patterns and health have not properly controlled for the confounding effects of physical activity, which, without exception, have been measured by self-report. Validation studies have consistently found that self-reported physical activity measures are biased (e.g. most people overestimate their activity) and only modestly associated with objective measures of physical activity. 2–4 The imprecision of these estimates would result in an underestimated effect of physical activity on the morbidity and mortality outcomes, and in studies of unhealthy dietary patterns and poor health, would result in residual confounding for physical activity. Our study attempted to overcome this limitation by including an objective marker of physical activity as a covariate in the analyses. Indeed, the risk estimates for dietary patterns were substantively smaller when cardiorespiratory fitness was the covariate compared with when self-reported physical activity was the covariate. Next, we would like to address the concern regarding the use of cardiorespiratory fitness as a marker of physical activity participation. We appreciate that fitness is not a direct measure of physical activity and that other factors such as genetics, age and sex play a role in determining one’s fitness. However, by using age- and sex-specific cut-points to define the different fitness groups, our analyses accounted for some of the most meaningful non-activity determinants of fitness. Furthermore, several studies have shown that fitness is highly related to physical activity participation in recent months 2,5,6 and that fitness is responsive to changes in physical activity. 7–9 Thus, while we recognize that there is not a perfect relation between physical activity and fitness, it is clear that physical activity is a major driver of fitness. We, therefore, are confident that cardiorespiratory fitness can be used as a proxy and objective measure of physical activity. Both commentaries argued that diet is an important determinant of cardiorespiratory fitness. We do not support this position. Randomized controlled trials have clearly demonstrated that diet-induced weight loss is not associated with improvements in fitness

Cedric and Frank Garland 1 observed that metropolitan states in the south-west of the US had lower mortality rates from colon cancer in 1959–61, on average ~10 per 100 000 per year ageadjusted in white males, than metropolitan states in the northwest, on average ~15 per 100 000. They reported that the rank order coefficient of correlation between colon cancer mortality and solar irradiance across the metropolitan states was 0.7. It was 0.6 across non-metropolitan states and the same values were obtained when cancer mortality rates for US counties in 1950–69 were used instead of the whole-of-state values. They noted, from broad regional data, that the probably mainly south to north gradient in colon cancer mortality did not appear to correlate well with patterns of dietary meat, fruit, or vegetable intake but acknowledged that its correlationwith solar irradiance might still be due to confounding with some other dietary, behavioural, or, perhaps, genetic factor. Apperly 2 had reported an association between mortality from all cancers and solar irradiance across the states of the US 40 years earlier and suggested ‘it may be possible to reduce our cancer deaths by exposing suitable skin areas to sunlight . . . insufficient to cause skin cancer’. While the Garlands did not acknowledge this work they can be reasonably excused for not having uncovered it. Unsurprisingly, the Garlands advanced the possibility that photolytic production of vitamin D in the skin might explain the correlation between colorectal cancer mortality and solar irradiance. This is the major source of vitamin D for most human populations. They advanced an argument about antiinflammatory effects of calcium (the absorption of which is promoted by vitamin D) in the bowel mucosa and evidence of a protective effect of milk against colon cancer in one US study (milk was and is mandatorily fortified with vitamin D in the US) in support of this possibility. The Garlands and others published papers reporting on correlations in the US between solar irradiance and cancers of the breast, 3 prostate, 4 and ovary 5 over the subsequent decade or so. None of these, or the original paper on colon cancer, initially attracted much attention. Indeed, the International Agency for Research on Cancer’s 1992 monograph on the carcinogenic risk of solar and ultraviolet radiation to humans stated that ‘no adequate data were available to the Working Group’ when referring to all cancers other than skin, lip, and ocular cancers. 6

I was pleased to hear that the errors in the 95% confidence limits presented by Garfinkel et al. (1) to which I referred in my previous letter (2) arose from misapplying the program by Rothman and Boice (3) rather than from an error in the program itself, which would have had more alarming consequences. Although I am sorry that I falsely accused the program of being erroneous, it remains a matter of concern that incorrect data have been published. Elsewhere, I propose some tests and procedures to help avoid such errors appearing in published papers, errors that are, unfortunately, not that uncommon in the literature on environmental tobacco smoke (ETS) (4). I was extremely surprised to hear Rothman and Boice (3) suggest that I rely on statistical significance as a primary indicator of effect, since in my previous letter I had cited a paper of mine (5) that argued at length that the statistically significant but weak association between lung cancer and smoking by the husband does not provide convincing evidence of a true effect of ETS exposure on risk of lung cancer. The reasons for this are precisely those to which Rothman and Boice (3) refer, biases such as those arising from misclassification of exposure to active smoking and from confounding. While it is certainly true that the presence of a statistically significant association does

Summary The paper examines the role of education as a causal channel through which growing up poor affects the economic outcomes in adulthood in the European Union. We apply a potential outcomes approach to quantify those effects and we provide a sensitivity analysis on possible unobserved confounders, such as child ability. Our estimates indicate that being poor in childhood significantly decreases the level of income in adulthood and increases the average probability of being poor. Moreover, our results reveal a significant role of education in this intergenerational transmission. These results are particularly relevant for Mediterranean and central and eastern European countries.

Background Young people with low education have worse health than those with higher education. This paper examined the extent to which employment and income reduced the adverse effects of low education on mental health among people aged 20-35 years. Methods We used causal mediation analyses to estimate the total causal effect (TCE) of low education on mental health and to decompose the effect into the natural direct effect (NDE) and the natural indirect effect (NIE) through two mediators examined sequentially: employment (labour-force participation/occupation skill level) and income. Three waves of the Household, Income and Labour Dynamics in Australia (HILDA) survey (2012-14) were used to establish a temporal sequence between low education (not completing high school), mediators and mental health [the Mental Health Inventory (MHI-5)] among participants aged 20-35 years. Among those who were employed, we conducted further analyses examining the effect of job characteristics as a mediator of the relationship between low education and mental health. Results The TCE of low education on the MHI-5 was -3.61 [95% confidence interval (CI) -5.30 to -1.92]. The NIE through labour force status and occupational skill level was -1.09 (95% CI -2.29 to 0.10) and -1.49 (95% CI -2.79 to -0.19) through both labour-force status/occupational skill level and income, corresponding to a percentage mediated of 41%. Among the employed, education had a much smaller effect on the MHI-5. Conclusions Improving employment opportunities could reduce nearly half of the adverse effects of low education on the mental health of young people.

Background Occupational exposure to polycyclic aromatic hydrocarbons (PAHs) has been shown to be associated with lung cancer in various epidemiological studies in industries such as aluminium reduction/smelting, coal gasification, coke production, iron/steel foundries, coal/coke and related products and carbon/graphite electrodes production. Aims To update data on the association between PAH exposure and morbidity and mortality due to lung cancer among workers in different occupations, including smoking data. Methods A comprehensive literature search was conducted to retrieve relevant papers for meta-analysis. Cohort studies with standardized mortality ratios or standardized incidence ratios and calculated overall risk ratio with their corresponding 95% confidence intervals (CIs) were included in the analysis. Chi-square test for heterogeneity was used to evaluate the consistency of findings between the studies. Results A significant risk of lung cancer was observed among the coal/coke and related product industry 1.55 (95% CI 1.01-2.37) and the iron/steel foundry industry 1.52 (95% CI 1.05-2.20). There was a wide variation in smoking habits and PAHs exposure among studies. Conclusions Coal/coke industry and iron/steel industry workers showed a higher risk of lung cancer compared with other occupations exposed to PAHs. The confounding effects of smoking and individual exposure levels of PAH should be taken into account.

While a growing body of research is examining the impacts of prolonged occupational sitting on cardiovascular and other health risk factors, relatively little work has examined the effects of occupational standing. The objectives of this paper were to examine the relationship between occupations that require predominantly sitting and those that require predominantly standing and incident heart disease. A prospective cohort study combining responses to a population health survey with administrative health-care records, linked at the individual level, was conducted in Ontario, Canada. The sample included 7,320 employed labor-market participants (50% male) working 15 hours a week or more and free of heart disease at baseline. Incident heart disease was assessed using administrative records over an approximately 12-year follow-up period (2003–2015). Models adjusted for a wide range of potential confounding factors. Occupations involving predominantly standing were associated with an approximately 2-fold risk of heart disease compared with occupations involving predominantly sitting. This association was robust to adjustment for other health, sociodemographic, and work variables. Cardiovascular risk associated with occupations that involve combinations of sitting, standing, and walking differed for men and women, with these occupations associated with lower cardiovascular risk estimates among men but elevated risk estimates among women.

Background: Self-reported health status (SRHS) reflects an individual’s perception of their social, biological and psychological health, and has been linked to increased mortality risk and increased use of health services. Having a psychological co-morbidity can reduce health outcomes and increase healthcare costs. This paper investigates the relationship between SRHS and signs of psychological distress (PD) in European urban settings. Methods: The study sample comprised 20,439 adult respondents to surveys conducted across 37 urban areas. Data on SRHS, signs of PD and potential confounders were analysed in a multivariable logistic regression. Results: After adjusting for potential confounders (gender, age, time lived in country and regularity of contact with family and friends), a significant association was found between PD and self-reported poor health for all urban areas, particularly in Western European countries (adjusted OR=3.615, 95% CI: 3.333-3.920 p

A critical assumption of causal inference is that of no unmeasured confounding: for estimated exposure effects to have valid causal interpretations, a sufficient set of predictors of exposure and outcome must be adequately measured and correctly included in the respective inference model(s). In an observational study setting, this assumption will often be unsatisfied, and the potential impact of unmeasured confounding on effect estimates should be investigated. The confounding function approach allows the impact of unmeasured confounding on estimates to be assessed, where unmeasured confounding may be due to unmeasured confounders and/or biases such as collider bias or information bias. Although this approach is easy to implement and pertains to the sum of all bias, its use has not been widespread, and discussion has typically been limited to continuous outcomes. In this paper, we consider confounding functions for use with binary outcomes and illustrate the approach with an example. We note that confounding function choice encodes assumptions about effect modification: some choices encode the belief that the true causal effect differs across exposure groups, whereas others imply that any difference between the true causal parameter and the estimate is entirely due to imbalanced risks between exposure groups. The confounding function approach is a useful method for assessing the impact of unmeasured confounding, in particular when alternative approaches, e.g. external adjustment or instrumental variable approaches, cannot be applied. We provide Stata and R code for the implementation of this approach when the causal estimand of interest is an odds or risk ratio.

Summary In both observational studies and randomized trials with noncompliance, unmeasured confounding may exist which may bias treatment effect estimates. Instrumental variables (IV) are a popular technique for addressing such confounding, enabling consistent estimation of causal effects. This paper proposes nonparametric IV estimators for censored time to event data that may be subject to competing risks. A simple, plug-in estimator is introduced using nonparametric estimators of the cumulative incidence function, with confidence intervals derived using asymptotic theory. To provide an overall test of the treatment effect, an integrated weighted difference statistic is suggested, which is applicable to data with and without competing risks. Simulation studies demonstrate that the methods perform well with realistic samples sizes. The methods are applied to assess the effect of infant or maternal antiretroviral therapy on transmission of HIV from mother to child via breastfeeding using data from a large, recently completed randomized trial in Malawi where noncompliance with assigned treatment may confound treatment effect estimates.

Summary Weighting methods offer an approach to estimating causal treatment effects in observational studies. However, if weights are estimated by maximum likelihood, misspecification of the treatment assignment model can lead to weighted estimators with substantial bias and variance. In this paper, we propose a unified framework for constructing weights such that a set of measured pretreatment covariates is unassociated with treatment assignment after weighting. We derive conditions for weight estimation by eliminating the associations between these covariates and treatment assignment characterized in a chosen treatment assignment model after weighting. The moment conditions in covariate balancing weight methods for binary, categorical and continuous treatments in cross-sectional settings are special cases of the conditions in our framework, which extends to longitudinal settings. Simulation shows that our method gives treatment effect estimates with smaller biases and variances than the maximum likelihood approach under treatment assignment model misspecification. We illustrate our method with an application to systemic lupus erythematosus data.

Researchers conducting observational studies need to consider 3 types of biases: selection bias, information bias, and confounding bias. A whole arsenal of statistical tools can be used to deal with information and confounding biases. However, methods for addressing selection bias and unmeasured confounding are less developed. In this paper, we propose general bounding formulas for bias, including selection bias and unmeasured confounding. This should help researchers make more prudent interpretations of their (potentially biased) results.

Missing data are a commonly occurring threat to the validity and efficiency of epidemiologic studies. Perhaps the most common approach to handling missing data is to simply drop those records with 1 or more missing values, in so-called “complete records” or “complete case” analysis. In this paper, we bring together earlier-derived yet perhaps now somewhat neglected results which show that a logistic regression complete records analysis can provide asymptotically unbiased estimates of the association of an exposure of interest with an outcome, adjusted for a number of confounders, under a surprisingly wide range of missing-data assumptions. We give detailed guidance describing how the observed data can be used to judge the plausibility of these assumptions. The results mean that in large epidemiologic studies which are affected by missing data and analyzed by logistic regression, exposure associations may be estimated without bias in a number of settings where researchers might otherwise assume that bias would occur.

Increasingly often in epidemiologic research, associations between survival time and predictors of interest are measured by differences between distribution functions rather than hazard functions. For example, differences in percentiles of survival time, expressed in absolute time units (e.g., weeks), may complement the popular risk ratios, which are unitless measures. When analyzing time to an event of interest (e.g., death) in prospective cohort studies, the time scale can be set to start at birth or at study entry. The advantages of one time origin over the other have been thoroughly explored for the estimation of risks but not for the estimation of survival percentiles. In this paper, we analyze the use of different time scales in the estimation of survival percentiles with Laplace regression. Using this regression method, investigators can estimate percentiles of survival time over levels of an exposure of interest while adjusting for potential confounders. Our findings may help to improve modeling strategies and ease interpretation in the estimation of survival percentiles in prospective cohort studies.

Summary More than 1100 abandoned mines, milling sites and waste piles from the uranium mining period are scattered across the Navajo Nation, resulting in exposures to environmental metals, including uranium. The Diné Network for Environmental Health project began in response to concerns regarding the community health effects of these environmental exposures on chronic disease. The paper presents the results of the initial Diné Network for Environmental Health survey of 1304 individuals living on the Navajo Nation. We examine the relationship between uranium mine waste exposure and kidney disease, diabetes and hypertension. These chronic diseases are found at high prevalences in the study population, present major public health risks and have been linked to metals exposures in other studies. We model the exposure–outcome relationship by using a multivariate model for the three binary responses. We implement a Bayesian multivariate t-model, which has marginal log-odds ratio parameter interpretations and is computationally efficient. In examining environmental exposures, appropriately adjusting for potential confounders is pivotal to obtaining policy relevant effect estimates. We use Bayesian model averaging to account for uncertainty in the functional form for confounding adjustment within a small set of measured confounders. Using this multivariate framework, we find evidence of associations between these chronic diseases and both historic mining era and legacy mining exposures.

A central question in causal inference with observational studies is the sensitivity of conclusions to unmeasured confounding. The classical Cornfield condition allows us to assess whether an unmeasured binary confounder can explain away the observed relative risk of the exposure on the outcome. It states that for an unmeasured confounder to explain away an observed relative risk, the association between the unmeasured confounder and the exposure, and also that between the unmeasured confounder and the outcome, must both be larger than the observed relative risk. In this paper, we extend the classical Cornfield condition in three directions. First, we consider analogous conditions for the risk difference, and allow for a categorical, not just a binary, unmeasured confounder. Second, we provide more stringent thresholds which the maximum of the above-mentioned associations must satisfy, rather than simply weaker conditions that both must satisfy. Third, we show that all previous results on Cornfield conditions hold under weaker assumptions than previously used. We illustrate their potential applications by real examples, where our new conditions give more information than the classical ones.

Summary The paper investigates two different multilevel approaches, the multilevel cross-classified and the multiple-membership models, for the analysis of interviewer effects on wave non-response in longitudinal surveys. The models proposed incorporate both interviewer and area effects to account for the non-hierarchical structure, the influence of potentially more than one interviewer across waves and possible confounding of area and interviewer effects arising from the non-random allocation of interviewers across areas. The methods are compared by using a data set: the UK Family and Children Survey.

Unobserved confounding can seldom be ruled out with certainty in nonexperimental studies. Negative controls are sometimes used in epidemiologic practice to detect the presence of unobserved confounding. An outcome is said to be a valid negative control variable to the extent that it is influenced by unobserved confounders of the exposure effects on the outcome in view, although not directly influenced by the exposure. Thus, a negative control outcome found to be empirically associated with the exposure after adjustment for observed confounders indicates that unobserved confounding may be present. In this paper, we go beyond the use of control outcomes to detect possible unobserved confounding and propose to use control outcomes in a simple but formal counterfactual-based approach to correct causal effect estimates for bias due to unobserved confounding. The proposed control outcome calibration approach is developed in the context of a continuous or binary outcome, and the control outcome and the exposure can be discrete or continuous. A sensitivity analysis technique is also developed, which can be used to assess the degree to which a violation of the main identifying assumption of the control outcome calibration approach might impact inference about the effect of the exposure on the outcome in view.

This paper reviews results of studies on effects of fetal programming and maternal nutrition during pregnancy on growth, efficiency, carcass, muscle, and meat quality characteristics of cattle. It includes results from our Australian Beef Cooperative Research Centre studies on factors such as chronic severe nutritional restriction from approximately d 80 of pregnancy to parturition and/or throughout lactation used to create early-life growth differences in the offspring of cows within pasture-based systems and the effect of these treatments on production characteristics to 30 mo of age. Fetal programming and related maternal effects are most pronounced and explain substantial amounts of variation for growth-related production characteristics such as BW, feed intake, carcass weight, muscle weights, meat yield, and fat and bone weights at any given age but are less evident when assessed at the same BW and carcass weight. Some effects of maternal and early-life factors in our studies were evident for efficiency traits but fewer affected beef quality characteristics at 30 mo of age, explaining only small amounts of variation in these traits. It is difficult to uncouple maternal nutritional effects specific to prenatal life from those that carry over to the postnatal period until weaning, particularly the effects of maternal nutrition during pregnancy on subsequent lactational performance. Hence, experimental design considerations for studying fetal programming effects on offspring during later life are discussed in relation to minimizing or removing prenatal and postnatal confounding effects. The relative contribution of fetal programming to the profitability of beef production systems is also briefly discussed. In this regard, the importance of health and survival of cows and calves, the capacity of cows to rebreed in a timely manner, and the efficiency with which feed and other resources are used cannot be overemphasized in relation to economics, welfare, and the environment.

Background Outpatient therapies for urinary tract infections (UTIs) are becoming limited due to antimicrobial resistance. The purpose of this paper is to report how the incidence of hospitalizations for UTIs have varied over time in both men and women and across age groups. We also explore how the severity for UTI hospitalizations has changed and describe the seasonality of UTI hospitalizations. Methods Using the Nationwide Inpatient Sample, we compute a time-series of UTI incidence and subdivide the series by age and sex. We fit a collection of time-series models to explore how the trend and seasonal intensity varies by age and sex. We modeled changes in severity using regression with available confounders. Results In 2011, there were approximately 400000 hospitalizations for UTIs with an estimated cost of $2.8 billion. Incidence increased by 52% between 1998 and 2011. The rate of increase was larger among both women and older patients. We found that the seasonal intensity (summer peaks and winter troughs) increased over time among women while decreasing among men. For both men and women, seasonality decreased with advancing age. Relative to controls and adjusted for demographics, we found that costs among UTI patients grew more slowly, patients left the hospital earlier, and patients had lower odds of death. Conclusions Incidence of UTI hospitalization is increasing and is seasonal, peaking in the summer. However, the severity of UTI admissions seems to be decreasing, indicating that patients previously treated as outpatients may now be admitted to the hospital due to increasing antimicrobial resistance.

Data regularly recorded in commercial herds have been used extensively for estimation of dis- ease incidence rates, for inferences regarding genetic and phenotypic associations between traits, or for devel- oping predictive models for economically important traits. Some studies have also used fi eld data to inves- tigate potential causal relationships between variables. However, inferring causal effects from observational data is complex due to potential confounding effects and careful analyses using specifi c statistical and data min- ing techniques as well as different sets of assumptions are required. Nonetheless, although virtually unknown in the agricultural research community, such methods are available and have been used in many other fi elds. In this paper, we review and discuss the analysis of obser- vational data using fi eld-recorded information and its potential utility in the study of causal effects in livestock. It is our postulation that there is much to be learned from such data, which can be used either to explicitly investigate causal relationships between variables or to generate hypotheses for further investigation using con- trolled experiments or additional fi eld-recorded data.

Summary The paper investigates the link between area-based socio-economic deprivation and the incidence of child pedestrian casualties. The analysis is conducted by using data for small spatial zones within major British cities over the period 2001–2007. Spatial longitudinal generalized linear mixed models, estimated by using frequentist and Bayesian approaches, are used to address issues of confounding, spatial dependence and transmission of deprivation effects across zones (i.e. interference). The results show a consistent strong deprivation effect across model specifications. The incidence of child pedestrian casualties in the most deprived zones is typically greater than 10 times that in the least deprived zones. Modelling interference through a spatially auto-regressive covariate uncovers a substantially larger effect.

Summary The paper describes a Bayesian spatial discrete time survival model to estimate the effect of air pollution on the risk of preterm birth. The standard approach treats prematurity as a binary outcome and cannot effectively examine time varying exposures during pregnancy. Time varying exposures can arise either in short-term lagged exposures due to seasonality in air pollution or long-term cumulative exposures due to changes in length of exposure. Our model addresses this challenge by viewing gestational age as time-to-event data where each pregnancy becomes at risk at a prespecified time (e.g. the 28th week). The pregnancy is then followed until either a birth occurs before the 37th week (preterm), or it reaches the 37th week, and a full-term birth is expected. The model also includes a flexible spatially varying baseline hazard function to control for unmeasured spatial confounders and to borrow information across areal units. The approach proposed is applied to geocoded birth records in Mecklenburg County, North Carolina, for the period 2001–2005. We examine the risk of preterm birth that is associated with total cumulative and 4-week lagged exposure to ambient fine particulate matter.