Search

In this last issue of 2008 I take the opportunity to report some observations on recent trends in articles published in the EJE. In the first decade after Antonio Sanna founded the Journal in 1985 the emphasis was clearly on infectious diseases. Over half of the papers were in this category, and the methods papers generally also were on infectious diseases. This gradually changed after Claude Hannoun took over as editor-in-chief in 1995. He published a growing proportion of articles on chronic diseases, in particular cardiovascular disease. In recent years, the Journal has published increasingly papers on epidemiologic and statistical methods, on the objectives and study design of new epidemiologic studies, on genetic epidemiology, and on developmental epidemiology. Here I will review some of the highlights of the articles published in recent years, including those on infectious diseases. Recent methods articles have addressed issues of measurement of determinants and outcome, exposuremeasurement error and information bias [1–10]. We have published some work on the design and evaluation of questionnaires, in particular on assessment of physical activity. We also have published important articles on causal modelling, on confounding, and on effect-modification and interaction. In our section devoted to new studies the majority op articles has been on new birth cohort studies and other studies addressing early-life causes of adult diseases, and on the biobanks organized as part of those efforts [11–17]. In genetic epidemiology we have published some metaanalyses of candidate gene studies on various diseases, and in the near future we will undoubtedly publish the results of the major genome-wide association studies that are currently underway in many epidemiologic research groups [18–22]. In addition, we have published commentaries on the value, of lack thereof, of predictive genetic testing for complex diseases and more in general on the promises of translational research in genomics. We have published a substantial number of studies in developmental epidemiology, addressing fetal and other early life determinants of adult diseases [23–34]. This involved in particular studies of low birth weight and cardiovascular disease and diabetes, in different ethnic populations, and under different nutritional conditions. Increasingly, hemodynamic and endocrinological studies have addressed the hypothesis of early life causes of adult diseases. We have also published on the new studies in this field and on the epidemiologic resources that are now being build-up for future developmental epidemiologic studies. The Journal publishes still a substantial number of articles on infectious diseases, about ten per cent of all published papers [35–55]. The largest number of articles in recent years has been on HIV/AIDS and on hepatitis, and quite often on their interrelations. We have also published work on new methods in infectious disease epidemiology, and on their applications to diseases like tuberculosis, malaria and meningitis. We will continue to emphasize the publication of important epidemiologic studies on infectious diseases in the Journal. The EJE has become more selective over the years. In the first decade of its existence about half of the manuscripts that came to the office were published. In recent years this has gone down to about 15%, reflecting both a substantially larger number of manuscripts received and a stricter acceptance policy. In this the editors have been strongly supported by the work for the Journal of many A. Hofman (&) Department of Epidemiology, Erasmus Medical Center, P.O. Box 2040, 3000 CA Rotterdam, The Netherlands e-mail: a.hofman@erasmusmc.nl

The ‘omics’-era is well and truly upon us. Driven by technological innovations, fields such as genetics, epigenetics, proteomics and imaging are advancing at tremendous pace and in addition to unraveling pathophysiological mechanisms of health and disease, also reveal the increasingly complex biology that forms the basis of living systems. In this issue of the European Journal of Epidemiology, Nishihara and colleagues describe ‘molecular pathological epidemiology’, a relatively novel sub-field within epidemiology that examines the association between a risk factor and disease by taking into account the inherent pathological (and often molecular) heterogeneity of a disease [1]. Molecular pathological epidemiology had its original applications in oncology, but—driven by omics-research—is now quickly disseminating to other fields of biomedical research. Molecular pathological epidemiology holds great promise by bringing together epidemiology and molecular biology, two fields that traditionally have been considered at the opposite extremes of the biomedical research spectrum. It is expected that in coming years molecular pathological epidemiology will importantly shape the way, in which omics-research is conducted and interpreted. In their paper, Nishihara et al. [1] use molecular pathological epidemiology to elucidate biomedical paradoxes that hitherto were ascribed to methodological phenomena, such as selection bias, collider bias, unmeasured confounding and reverse causality. They show that paradoxical associations that have been found between a risk factor for disease and prognosis after disease-onset might in fact be reflecting true underlying biology rather than being an artifact of study design. While the focus of their paper is on examples that elegantly illustrate the application of molecular pathological epidemiology to unravel methodological paradoxes, I believe this approach will have further-reaching implications in biomedical research, which are only briefly touched upon by Nishihara et al. and which I will expand on below. First, I am convinced that molecular pathological epidemiology is a pivotal cornerstone for successful implementation of novel initiatives, such as ‘personalized medicine’ and its latest incarnation ‘precision medicine’. Precision medicine is probably an overhyped concept at the moment, but nonetheless does signal the direction in which clinical medicine should develop to reach the next level. The basic concept of precision medicine is that each individual has his or her own unique set of characteristics that together cause disease or determine prognosis. This unique set therefore requires a unique approach to clinical management, tailor-made for that individual patient [2]. The continued success of omics-research through novel discoveries will lead to rapid expansion of the known set of characteristics that determine disease and thereby increase the precision of precision medicine. Interestingly, this framework for disease causation was formalized in the epidemiological literature already as far back as the seventies [3]. As such, precision medicine is nothing new for epidemiology. In fact, the enthusiasm with which precision medicine has been heralded by various stake-holders and policy makers could and should be considered a vindication for the epidemiological way of thinking about disease. What is worrying though is that the role of the epidemiologist in precision medicine seems so far to be ill-defined. Yet, given the challenges that precision medicine faces for correct design and interpretation of its research (indeed, consider the paradoxical findings that can only be tackled by proper understanding of molecular pathological epidemiology and other epidemiological concepts), I would expect the epidemiologist to be at the fore-front of precision medicine. Therefore, now is the time for epidemiologists to assume a pro-active role in the process of developing and implementing precision medicine. The second implication of molecular pathological epidemiology relates to a core activity in clinical medicine. Diagnosis of disease (and the efforts to achieve that) forms a major milestone in the clinical management of any patient and in fact much—if not most—effort in clinical practice is dedicated at establishing an accurate diagnosis. Diagnoses of diseases are usually based on a set of signs and symptoms that typically occur together in a patient and as such are considered a separate disease entity. Very often, diagnostic criteria also include the presumed underlying etiology, thereby forming the basis of etiologic subtyping. Examples include subtyping of ischemic stroke into large vessel, small vessel, or cardioembolic stroke and polyneurpathy into various subgroups, such as diabetic polyneuropathy. Such subclassifications have been developed for several decades and though very useful for contemporary clinical management, neglect the multifactorial nature of diseases [2]. Consequently, the underlying presumed etiology is neither sensitive nor specific enough to justify a single dedicated subcategory. For instance, most stroke patients will have a certain degree of small vessel pathology and it therefore remains arbitrary what burden of small vessel pathology warrants the designation of a stroke as small vessel stroke. Molecular pathological epidemiology can provide an etiologic framework for diseases, which does allow for etiologic subtyping. Taking examples from Nishihara et al. it can be justified to subclassify colorectal into two types based on presence or absence of the rs4939837 variant in the SMAD7-gene or subtyping renal cell carcinoma based on up- or downregulation of fatty acid synthase. After all, rs4939837 and up- or downregulation of fatty acide synthase may act as criteria sensitive and specific enough to distinguish the two subtypes. Omics-research will continue to further unravel many complex diseases and therefore expand the possibilities for etiological subtyping. Therefore, now is the time to re-write medical textbooks and incorporate molecular pathological epidemiology into mainstream biomedical literature. Just like any other field, epidemiology too needs to keep up with the technology-push in research. Molecular pathological epidemiology might just be the trigger that helps epidemiology firmly establish its role in the omics-era.

This paper examines whether the anomalous summer peak in deaths from coronary heart disease (CHD) in Finland could be attributed to adverse effects of the Midsummer festival and alcohol consumption during the festival. Daily deaths from CHD and alcohol poisoning in Finland, 1961–2014, that occurred during the 7 days centering on Midsummer Day were analysed in relation to deaths during 14 to 4 days before and 4 to 14 after Midsummer Day. Daily counts of deaths from CHD among persons aged 35–64 years were regressed on days around the Midsummer period by negative binomial regression. Mortality from CHD was highest on Midsummer Day (RR 1.25 (95% confidence interval 1.12–1.31), one day after the peak in deaths from alcohol poisonings. RR for CHD on Midsummer Day was particulary high (RR = 1.43; 1.09–1.86) in the 2000s, 30% of deaths being attributable to that day. In conclusion, the anomalous and prominent summer peak in deaths from CHD in Finland is an adverse consequence of the Midsummer festival. The most likely underlying reason is heavy alcohol consumption during the festival period, especially on Midsummer Eve. In the 2000s, one third of deaths from CHD on Midsummer Day are preventable.

The nine Bradford Hill (BH) viewpoints (sometimes referred to as criteria) are commonly used to assess causality within epidemiology. However, causal thinking has since developed, with three of the most prominent approaches implicitly or explicitly building on the potential outcomes framework: directed acyclic graphs (DAGs), sufficient-component cause models (SCC models, also referred to as ‘causal pies’) and the grading of recommendations, assessment, development and evaluation (GRADE) methodology. This paper explores how these approaches relate to BH’s viewpoints and considers implications for improving causal assessment. We mapped the three approaches above against each BH viewpoint. We found overlap across the approaches and BH viewpoints, underscoring BH viewpoints’ enduring importance. Mapping the approaches helped elucidate the theoretical underpinning of each viewpoint and articulate the conditions when the viewpoint would be relevant. Our comparisons identified commonality on four viewpoints: strength of association (including analysis of plausible confounding); temporality; plausibility (encoded by DAGs or SCC models to articulate mediation and interaction, respectively); and experiments (including implications of study design on exchangeability). Consistency may be more usefully operationalised by considering an effect size’s transportability to a different population or unexplained inconsistency in effect sizes (statistical heterogeneity). Because specificity rarely occurs, falsification exposures or outcomes (i.e., negative controls) may be more useful. The presence of a dose-response relationship may be less than widely perceived as it can easily arise from confounding. We found limited utility for coherence and analogy. This study highlights a need for greater clarity on BH viewpoints to improve causal assessment.

In this paper we study approaches for dealing with treatment when developing a clinical prediction model. Analogous to the estimand framework recently proposed by the European Medicines Agency for clinical trials, we propose a `predictimand' framework of different questions that may be of interest when predicting risk in relation to treatment started after baseline. We provide a formal definition of the estimands matching these questions, give examples of settings in which each is useful and discuss appropriate estimators including their assumptions. We illustrate the impact of the predictimand choice in a dataset of patients with end-stage kidney disease. We argue that clearly defining the estimand is equally important in prediction research as in causal inference., Comment: under review

Childhood cancer survivors face risks from a variety of late effects, including cardiac events, second cancers, and late mortality. The aim of the pan-European PanCare Childhood and Adolescent Cancer Survivor Care and Follow-Up Studies (PanCareSurFup) Consortium was to collect data on incidence and risk factors for these late effects among childhood cancer survivors in Europe. This paper describes the methodology of the data collection for the overall PanCareSurFup cohort and the outcome-related cohorts. In PanCareSurFup 13 data providers from 12 countries delivered data to the data centre in Mainz. Data providers used a single variable list that covered all three outcomes. After validity and plausibility checks data was provided to the outcome-specific working groups. In total, we collected data on 115,596 patients diagnosed with cancer from 1940 to 2011, of whom 83,333 had survived 5 years or more. Due to the eligibility criteria and other requirements different numbers of survivors were eligible for the analysis of each of the outcomes. Thus, 1014 patients with at least one cardiac event were identified from a cohort of 39,152 5-year survivors; for second cancers 3995 survivors developed at least one second cancer from a cohort of 71,494 individuals, and from the late mortality cohort of 79,441 who had survived at least 5 years, 9247 died subsequently. Through the close cooperation of many European countries and the establishment of one central data collection and harmonising centre, the project succeeded in generating the largest cohort of children with cancer to date.

Neuroimaging plays an important role in etiologic research on neurological diseases in the elderly. The Rotterdam Scan Study was initiated as part of the ongoing Rotterdam Study with the aim to unravel causes of neurological disease by performing neuroimaging in a population-based longitudinal setting. In 1995 and 1999 random subsets of the Rotterdam Study underwent neuroimaging, whereas from 2005 onwards MRI has been implemented into the core protocol of the Rotterdam Study. In this paper, we discuss the background and rationale of the Rotterdam Scan Study. We also describe the imaging protocol and post-processing techniques, and highlight the main findings to date. Finally, we make recommendations for future research, which will also be the main focus of investigation in the Rotterdam Scan Study.