Standard errors for attributable risk for simple and complex sample designs

Adjusted attributable risk (AR) is the proportion of diseased individuals in a population that is due to an exposure. We consider estimates of adjusted AR based on odds ratios from logistic regression to adjust for confounding. Influence function methods used in survey sampling are applied to obtain simple and easily programmable expressions for estimating the variance of AR. These variance estimators can be applied to data from case-control, cross-sectional, and cohort studies with or without frequency or individual matching and for sample designs with subject samples that range from simple random samples to (sample) weighted multistage stratified cluster samples like those used in national household surveys. The variance estimation of AR is illustrated with: (i) a weighted stratified multistage clustered cross-sectional study of childhood asthma from the Third National Health and Examination Survey (NHANES III), and (ii) a frequency-matched case-control study of melanoma skin cancer.     

Attributable risk function in the proportional hazards model for censored time-to-event

Time-to-event endpoints are often used in clinical and epidemiological studies to evaluate disease association with hazardous exposures. In the statistical literature of time-to-event analysis, such association is usually measured by the hazard ratio in the proportional hazards model. In public health, it is also of important interest to assess the excess risk attributable to an exposure in a given population. In this article, we extend the notion of 'population attributable fraction' for the binary outcomes to the attributable risk function for the event times in prospective studies. A simple estimator of the time-varying attributable risk function is proposed under the proportional hazards model. Its inference procedures are established. Monte-Carlo simulation studies are conducted to evaluate its validity and performance. The proposed methodology is motivated and demonstrated by the data collected in a multicenter acquired immunodeficiency syndrome (AIDS) cohort study to estimate the attributable risk of human immunodeficiency virus type 1 (HIV-1) infections due to several potential risk factors.     

Proportion of neural tube defects attributable to known risk factors

BACKGROUND

Recognized risk factors for neural tube defects (NTDs) poorly predict population‐level NTD risk. However, the proportion of NTDs that can be attributed to these risk factors is uncertain.

METHODS

To determine the proportion of NTD cases that is attributable to known or suspected risk factors (i.e., female infant sex, family history of NTDs, and maternal Hispanic ethnicity, obesity, pregestational diabetes, gestational diabetes, low dietary folate intake, lack of folic acid supplementation, anticonvulsant use, and hot tub or sauna use), we estimated the adjusted population attributable fraction (aAF) for each factor, using the method of Eide and Geffler and data from the National Birth Defects Prevention Study.

RESULTS

Our analyses of these data indicate that the proportion of cases of spina bifida and anencephaly that can be attributed to known risk factors is 28% and 44%, respectively. For spina bifida, the factor with the greatest attributable fraction was maternal obesity (aAF, 10%), whereas for anencephaly it was Hispanic ethnicity (aAF, 15%).

CONCLUSION

Our analyses indicate that known risk factors account for <50% of NTD cases. Hence, the majority of NTD cases are attributable to, as yet, unidentified factors. These findings highlight the need for continued research to identify genetic and additional nongenetic risk factors for NTDs. Further, these findings suggest that strategies that aim to reduce the risk of NTDs associated with maternal Hispanic ethnicity and obesity may have the greatest impact on the population prevalence of these conditions. Birth Defects Research (Part A), 2013. © 2012 Wiley Periodicals, Inc.     

An Annotated Bibliography on the Attributable Risk

The attributable risk constitutes an important epidemiologic risk measure. In epidemiologic studies it quantifies the proportion of cases of disease due to the exposure factor under study. Hence, it assesses the public health importance of an exposure factor in the study population. The literature on the concept of attributable risk is diverse, appears in a variety of journals from different specialities, and most papers cite only a very limited part of it. In addition, it suffers from a substantial confusion in terminology and algebraic formulation. As a consequence, it is difficult to obtain an overview with respect to the statistical results relating to the concept. To remedy this problem, this paper provides an annotated bibliography containing a complete list of all references dealing with methodological aspects of the attributable risk.     

Estimation of population attributable fractions from fitted incidence ratios and exposure survey data, with an application to electromagnetic fields and childhood leukemia

Standard presentations of epidemiological results focus on incidence-ratio estimates derived from regression models fit to specialized study data. These data are often highly nonrepresentative of populations for which public-health impacts must be evaluated. Basic methods are provided for interval estimation of attributable fractions from model-based incidence-ratio estimates combined with independent survey estimates of the exposure distribution in the target population of interest. These methods are illustrated in estimation of the potential impact of magnetic-field exposures on childhood leukemia in the United States, based on pooled data from 11 case-control studies and a U.S. sample survey of magnetic-field exposures.     

Chewing disability in older adults attributable to tooth loss and other oral conditions

doi: 10.1111/j.1741‐2358.2010.00412.x Chewing disability in older adults attributable to tooth loss and other oral conditions Background: This study evaluates associations between oral health‐related factors and chewing ability, and quantifies the risk contributed by each factor. Materials and methods: Chewing ability and information on number of teeth, dentures and dental problems over the last 12 months were collected by mailing questionnaires to a random sample of 60‐ to 71‐year‐olds from Adelaide, South Australia. Logistic regression was used to model oral status and oral symptoms as predictors of chewing disability, and to estimate the population‐attributable fraction. Results: A total of 444 persons responded (response rate = 68.8%). Among dentate subjects, 10.3% were chewing‐deficient, with chewing disability more prevalent (p < 0.05) among those with <21 teeth (26.4%), dentures (20.4%), painful aching in the mouth (25.4%), pain in the face (16.7%), broken/chipped teeth (15.6%), sensitive teeth (14.1%), loose teeth (37.1%), and sore gums (18.0%). Adjusted Odds ratios (OR) showed inadequate dentition (OR = 4.20), painful aching in the mouth (OR = 4.88), and presence of loose teeth (OR = 4.70) were associated with chewing disability (p < 0.01), and their population attributable fractions were 18.5%, 15.1% and 7.8% respectively. Conclusions: Loose teeth, number of teeth and pain in the mouth were associated with chewing disability, with an inadequate dentition and pain in the mouth contributing most to chewing disability in this population.     

A scaled sample space cube used to illustrate attributable fractions

A three-dimensional graphic method is proposed for displaying the association structure between multiple explanatory variables and their relation to a categorical response. The method combines the techniques of mosaic displays and scaled Venn diagrams, and is especially useful for illustrating attributable fractions in epidemiology. The primary purpose is to show the reduction of disease risk in a population if the joint exposure distribution or the conditional risk function is modified, and the method can be extended to illustrate the potential effects of successive removal of exposures on the overall risk of disease. The scaled sample space cube may be used for communicating the difficult concept of attributable fraction to statisticians, the medical community and the general public in an easily understandable way. Demonstrations of the method use theoretical models as well as data from the Hordaland study on the effect of smoking and occupational exposure on obstructive lung disease. Also, the general principle of adding a third dimension to a mosaic display, instead of using shading or colouring, to show an attribute of a cell in a multiway contingency table, can be helpful for other purposes, as in residual analysis of a loglinear model fitting.     

Comparison of confidence intervals for adjusted attributable risk estimates under multinomial sampling

The epidemiologic concept of the adjusted attributable risk is a useful approach to quantitatively describe the importance of risk factors on the population level. It measures the proportional reduction in disease probability when a risk factor is eliminated from the population, accounting for effects of confounding and effect-modification by nuisance variables. The computation of asymptotic variance estimates for estimates of the adjusted attributable risk is often done by applying the delta method. Investigations on the delta method have shown, however, that the delta method generally tends to underestimate the standard error, leading to biased confidence intervals. We compare confidence intervals for the adjusted attributable risk derived by applying computer intensive methods like the bootstrap or jackknife to confidence intervals based on asymptotic variance estimates using an extensive Monte Carlo simulation and within a real data example from a cohort study in cardiovascular disease epidemiology. Our results show that confidence intervals based on bootstrap and jackknife methods outperform intervals based on asymptotic theory. Best variants of computer intensive confidence intervals are indicated for different situations.     

Cancers in France in 2015 attributable to high body mass index

BackgroundOverweight, as defined by high body mass index (BMI), is an established risk factor for various morbidities including cancer. Globally, its prevalence has increased markedly over the past decades. The aim of this study was to estimate the proportion and number of cancers that were attributable to high BMI in France in 2015.MethodsPopulation attributable fractions (PAFs) and numbers of cancer cases attributable to high BMI (a population mean BMI above the optimum of 22 kg/m²) were estimated by age and sex, for cancer sites with convincing or probable evidence of an established causal link. Assuming a 10-year lag-period, PAFs were calculated using mean BMI estimates from a cross-sectional French population survey, and relative risk estimates from published meta-analyses.ResultsAn estimated 18,639 cancer cases diagnosed in France in 2015 were attributable to high BMI, corresponding to 5.3% of all cancer cases (6.7% in women and 4.1% in men). This included 4507 cases of postmenopausal breast and 3380 cases of colon cancer. The highest estimated PAFs were for oesophageal adenocarcinoma and corpus uteri cancer (37% and 34%, respectively).ConclusionHigh BMI is associated with a substantial number of cancer cases in France, a country with a low but increasing prevalence of overweight and obesity when compared to other European countries. Assuming that the association between high BMI and cancer is causal, these results highlight the need to prioritise the prevention of this risk factor as part of cancer control planning in France and elsewhere in Europe.     

Cancers in France in 2015 attributable to insufficient physical activity

IntroductionInsufficient physical activity is a known risk factor for various co-morbidities, including cancer. Globally, its prevalence has increased markedly over the past decades. The aim of this study was to estimate the proportion and number of cancers that were attributable to insufficient physical activity in France in 2015.MethodsPopulation attributable fractions (PAFs) and numbers of cancer cases attributable to insufficient physical activity (<30 min daily of moderate-to-vigorous physical activity) were estimated by age, sex and cancer site. Assuming a 10-year lag-period, PAFs were calculated using physical activity prevalence from a cross-sectional French population survey and cancer-specific relative risks.ResultsAbout half of all French adults were found to be insufficiently physically active, with great variation by age and sex. In 2015, an estimated 2973 cancer cases diagnosed in French adults aged 30y+ were attributable to insufficient physical activity, corresponding to 0.8% of all cancer cases (0.2% in men and 1.6% in women). This comprised 3.8% of all postmenopausal breast cancers (1620 cases), 3.6% of all colon cancers (902 cases) and 6.0% of all cancers of the corpus uteri (450 cases). If at least half of the recommended physical activity level was achieved, 1095 cancer cases could have been avoided.ConclusionInsufficient physical activity is associated to about 3000 cancer cases in France, a country with comparatively low but increasing prevalence of this risk factor. This result is important for setting priorities in cancer prevention programmes aiming to increase physical activity in France and Europe in general.     

Sex differences in the proportion of esophageal squamous cell carcinoma cases attributable to tobacco smoking and alcohol consumption

Objective: Alcohol and tobacco are the two major established environmental factors associated with squamous cell carcinoma of the esophagus (ESCC). However, the prevalence of these exposures differs substantially between men and women. Moreover, the prevalence of smoking has declined in recent years, whereas per capita consumption of alcohol has remained steady in both sexes. Quantifying the burden of ESCC attributable to these causal factors is necessary to inform potential preventive strategies. Methods: We estimated the population attributable fraction (PAF) of ESCC due to smoking and alcohol, using data from an Australian population based case–control study (305 ESCC cases, 1554 controls). Results: Estimated PAF for ESCC were 49% (95% CI: 38–60) and 32% (95% CI: 25–40) due to smoking and heavy alcohol consumption respectively. More than 75% of the ESCC burden in men could be attributed to smokers with heavy alcohol consumption. The highest burden was among ≥30 pack years smokers who also consumed alcohol heavily (>17 drinks/week); this differed significantly between men (PAF 36%, 95% CI 29–44) and women (PAF 5%, 95% CI 2–10). Among women only, low intakes of fruit and vegetables accounted for about 9% of the ESCC burden. Conclusion: The burden of ESCC attributable to smoking combined with heavy alcohol consumption is remarkably high in men. In women, the burden of ESCC due to these factors is lower, and poor nutrition may also play a role.     

Estimation of cancer mortality attributable to excess body weight during 2006–2015 in China

BackgroundCancer is one of the most common causes of death. Excess body weight (EBW), a risk factor for cancer, is highly prevalent in China. We aimed to estimate the number and proportion of cancer deaths attributed to EBW and their changes during 2006–2015 in China.MethodsPopulation attributable fractions in 2006, 2010, and 2015 were calculated with 1) prevalence of overweight/obesity, exacted from the China Health and Nutrition Survey conducted in 8–9 provinces of China in 1997, 2000, and 2004; 2) relative risks for EBW and site-specific cancers, obtained from previous studies; 3) data on cancer deaths in 2006, 2010, and 2015, originated from the Chinese Cancer Registry Annual Report.ResultsIn 2015, EBW contributed to 45,918 (3.1% of all) cancer deaths in China, with 24,978 (2.6%) in men and 20,940 (3.8%) in women. By region, the fraction of cancer deaths attributable to EBW ranged from 1.6% (West) to 4.1% (Northeast). Cancers of liver, stomach, and colorectum were the main EBW-attributable cancers. The fractions of cancer deaths attributable to EBW were 2.4% (95%CI: 0.8–4.2%) in 2006, 2.9% (95%CI: 1.0–5.2%) in 2010, and 3.1% (95%CI: 1.0–5.4%) in 2015, respectively, and increased for all gender, region, and cancer site during 2006–2015.ConclusionsThe proportion of cancer deaths attributed to EBW was higher in women and Northeastern China, with an upward trend in the recent decade. A combination of comprehensive and individualized measures is necessary to reduce the prevalence of EBW and related cancer burden in China.     

On the conversion of solid cancer excess relative risk into lifetime attributable risk

Risk coefficients representing the lifetime radiation-induced cancer mortality (or incidence) attributable to an exposure to ionizing radiation, have been published by major international scientific committees. The calculations involve observations in an exposed population and choices of a standard population (for risk transportation), of suitable numerical models, and of computational techniques. The present lack of a firm convention for these choices makes it difficult to inter-compare risk estimates presented by different scientific bodies. Some issues that relate to a necessary harmonization and standardization of risk estimates are explored here. Computational methods are discussed and, in line with the approach utilized by ICRP, conversion factors from excess relative risk (ERR) to lifetime attributable risk (LAR) are exemplified for exposures at all ages and for occupational exposures. A standard population is specified to illustrate the possibility of a simplified standard for risk transportation computations. It is suggested that a more realistic perception of lifetime risk could be gained by the use of coefficients scaled to the lifetime spontaneous cancer rates in the standard population. The resulting quantity lifetime fractional risk (LFR) is advantageous also because it depends much less on the choice of the reference population than the lifetime attributable risk (LAR). Received: 5 April 2001 / Accepted: 1 July 2001     

Individual risk perception and empirical social structures shape the dynamics of infectious disease outbreaks

The dynamics of a spreading disease and individual behavioral changes are entangled processes that have to be addressed together in order to effectively manage an outbreak. Here, we relate individual risk perception to the adoption of a specific set of control measures, as obtained from an extensive large-scale survey performed via Facebook—involving more than 500,000 respondents from 64 countries—showing that there is a “one-to-one” relationship between perceived epidemic risk and compliance with a set of mitigation rules. We then develop a mathematical model for the spreading of a disease—sharing epidemiological features with COVID-19—that explicitly takes into account non-compliant individual behaviors and evaluates the impact of a population fraction of infectious risk-deniers on the epidemic dynamics. Our modeling study grounds on a wide set of structures, including both synthetic and more than 180 real-world contact patterns, to evaluate, in realistic scenarios, how network features typical of human interaction patterns impact the spread of a disease. In both synthetic and real contact patterns we find that epidemic spreading is hindered for decreasing population fractions of risk-denier individuals. From empirical contact patterns we demonstrate that connectivity heterogeneity and group structure significantly affect the peak of hospitalized population: higher modularity and heterogeneity of social contacts are linked to lower peaks at a fixed fraction of risk-denier individuals while, at the same time, such features increase the relative impact on hospitalizations with respect to the case where everyone correctly perceive the risks.     

The 'common disease-common variant' hypothesis and familial risks

The recent large genotyping studies have identified a new repertoire of disease susceptibility loci of unknown function, characterized by high allele frequencies and low relative risks, lending support to the common disease-common variant (CDCV) hypothesis. The variants explain a much larger proportion of the disease etiology, measured by the population attributable fraction, than of the familial risk. We show here that if the identified polymorphisms were markers of rarer functional alleles they would explain a much larger proportion of the familial risk. For example, in a plausible scenario where the marker is 10 times more common than the causative allele, the excess familial risk of the causative allele is over 10 times higher than that of the marker allele. However, the population attributable fractions of the two alleles are equal. The penetrance mode of the causative locus may be very difficult to deduce from the apparent penetrance mode of the marker locus.     

Burden of cancer attributable to tobacco smoking in member countries of the Association of Southeast Asian Nations (ASEAN), 2012

BackgroundCancer is an increasing problem in ASEAN (Association of Southeast Asian Nations). Tobacco use is a well-established risk factor for many types of cancers. Evidence on burden of cancer attributable to tobacco is essential to raise public and political awareness of the negative effects of tobacco on cancer and to be used to stimulate political action aims at reducing smoking prevalence in ASEAN member countries. The objective of this study was to estimate burden of cancer attributable to tobacco smoking in ASEAN, 2012.MethodsIn this study, smoking prevalence was combined with Relative Risks (RRs) of cancer to obtain Smoking Attributable Fractions (SAFs). Cancer incidence and mortality data among individuals aged 15 years and older were derived from GLOBOCAN 2012. Fourteen types of cancer were included in the analysis. Sensitivity analyses were conducted to examine the impact of the use of alternative RRs and the use of alternative prevalence of smoking in some countries.ResultsThe findings showed that tobacco smoking was responsible for 131,502 cancer incidence and 105,830 cancer mortality in ASEAN countries in 2012. In other words, tobacco smoking was accounted for 28.4% (43.3% in male and 8.5% in female) of cancer incidence and 30.5% (44.2% in male and 9.4% in female) of cancer mortality in ASEAN. When looking at the types of cancer, lung cancer showed the strongest association with tobacco smoking. Incidence of cancer and cancer mortality attributable to tobacco smoking varied by countries due to the differences in size of population, background risk of cancer, and prevalence of smoking in each country. According to the sensitivity analyses, RRs of lung cancer, pharynx cancer, and larynx cancer used in the estimates have significant impact on the estimates.ConclusionsAs about one-third of cancer incidence and mortality in ASEAN are attributable to tobacco smoking ASEAN member countries are strongly encouraged to put in place stronger tobacco control policies and to strengthen the existing tobacco control measure in order to effectively control cancer.     

Proportion of cancer cases and deaths attributable to lifestyle risk factors in Brazil

BackgroundLifestyle risk factors (tobacco smoking, alcohol consumption, overweight and obesity, unhealthy diet, and lack of physical activity) have been associated with increased risk of at least 20 types of cancer. We estimated the proportion of cancer cases and deaths that could be potentially avoided by eliminating or reducing lifestyle risk factors in Brazil.MethodsWe obtained the distribution of lifestyle risk factors by sex and age groups from recent representative health surveys in Brazil; relative risks from pooled analyses of prospective studies and meta-analyses; and cancer cases and deaths in 2012 from GLOBOCAN.ResultsWe found that 26.5% (114,497 cases) of all cancer cases and 33.6% (63,371 deaths) of all cancer deaths could be potentially avoided by eliminating lifestyle risk factors in Brazil. Plausible reductions in these exposures based on policy targets and cancer prevention recommendations could have potentially avoided 4.5% (19,731 cases) and 6.1% (11,480 deaths) of all cancer cases and deaths, respectively. Tobacco smoking accounted for most of the preventable cancer cases and deaths, followed by high body mass index and alcohol consumption. Larynx, lung, oropharynx, esophagus and colorectum cancer cases and deaths could be at least halved by eliminating these lifestyle risk factors.ConclusionFindings from this study may be useful to inform strategies for cancer prevention and control across Brazil.     

Tobacco and alcohol as risk factors for oesophageal cancer in a high incidence area in South Africa

BackgroundThe Eastern Cape Province of South Africa, which includes the former Transkei has high rates of squamous cell oesophageal cancer (OC), thought to be caused mainly by nutritional deficiencies and fungal contamination of staple maize. A hospital-based case-control study was conducted at three of the major referral hospitals in this region to measure, among other suspected risk factors, the relative importance of tobacco smoking and alcohol consumption for the disease in this population.MethodsIncident cases (n = 670) of OC and controls (n = 1188) were interviewed using a structured questionnaire which included questions on tobacco and alcohol-related consumption. Odds ratios (ORs) with 95% confidence intervals for each of the risk factors were calculated using unconditional multiple logistic regression models.ResultsA monotonic dose-response was observed across the categories of each tobacco-related variable in both sexes. Males and females currently smoking a total of >14 g of tobacco per day were observed to have over 4-times the odds of developing OC (males OR = 4.36, 95% CI 2.24–8.48; females OR = 4.56, 95% CI 1.46–14.30), with pipe smoking showing the strongest effect. Similar trends were observed for the alcohol-related variables. The quantity of ethanol consumed was the most important factor in OC development rather than any individual type of alcoholic beverage, especially in smokers. Males and females consuming >53 g of ethanol per day had approximately 5-times greater odds in comparison to non-drinkers (males OR = 4.72, 95% CI 2.64–8.41; females OR = 5.24, 95% CI 3.34–8.23) and 8.5 greater odds in those who smoked >14 g tobacco daily. The attributable fractions for smoking and alcohol consumption were 58% and 48% respectively, 64% for both factors combined.ConclusionTobacco and alcohol use are major risk factors for OC development in this region.ImpactThis study provides evidence for further reinforcement of cessation of smoking and alcohol consumption to curb OC development.     

Epidemiological evaluation of the use of genetics to improve the predictive value of disease risk factors.

The prevention of common diseases relies on identifying risk factors and implementing intervention in high-risk groups. Nevertheless, most known risk factors have low positive predictive value (PPV) and low population-attributable fraction (PAF) for diseases (e.g., cholesterol and coronary heart disease). With advancing genetic technology, it will be possible to refine the risk-factor approach to target intervention to individuals with risk factors who also carry disease-susceptibility allele(s). We provide an epidemiological approach to assess the impact of genetic testing on the PPV and PAF associated with risk factors. Under plausible models of interaction between a risk factor and a genotype, we derive values of PPV and PAF associated with the joint effects of a risk factor and a genotype. The use of genetic testing can markedly increase the PPV of a risk factor. PPV increases with increasing genotype-risk factor interaction and increasing marginal relative risk associated with the factor, but it is inversely proportional to the prevalences of the genotype and the factor. For example, for a disease with lifetime risk of 1%, if all the risk-factor effect is confined to individuals with a susceptible genotype, a risk factor with 10% prevalence and disease relative risk of 2 in the population will have a disease PPV of 1.8%, but it will have a PPV of 91.8% among persons with a genotype of 1% prevalence. On the other hand, genetic testing and restriction of preventive measures to those susceptible may decrease the PAF of the risk factor, especially at low prevalences of the risk factor and genotype.(ABSTRACT TRUNCATED AT 250 WORDS)     

The population-attributable fraction for time-dependent exposures using dynamic prediction and landmarking

The public health impact of a harmful exposure can be quantified by the population-attributable fraction (PAF). The PAF describes the attributable risk due to an exposure and is often interpreted as the proportion of preventable cases if the exposure was extinct. Difficulties in the definition and interpretation of the PAF arise when the exposure of interest depends on time. Then, the definition of exposed and unexposed individuals is not straightforward. We propose dynamic prediction and landmarking to define and estimate a PAF in this data situation. Two estimands are discussed which are based on two hypothetical interventions that could prevent the exposure in different ways. Considering the first estimand, at each landmark the estimation problem is reduced to a time-independent setting. Then, estimation is simply performed by using a generalized-linear model accounting for the current exposure state and further (time-varying) covariates. The second estimand is based on counterfactual outcomes, estimation can be performed using pseudo-values or inverse-probability weights. The approach is explored in a simulation study and applied on two data examples. First, we study a large French database of intensive care unit patients to estimate the population-benefit of a pathogen-specific intervention that could prevent ventilator-associated pneumonia caused by the pathogen Pseudomonas aeruginosa. Moreover, we quantify the population-attributable burden of locoregional and distant recurrence in breast cancer patients.